Research ArticleNonoverlay Heterogeneous Network Planning forEnergy Efficiency
Mahmut DemirtaG and Alkan Soysal
Electrical and Electronics Engineering Bahcesehir University Istanbul Turkey
Correspondence should be addressed to Alkan Soysal alkansoysalengbauedutr
Received 27 July 2016 Revised 27 December 2016 Accepted 22 January 2017 Published 9 February 2017
Academic Editor Simone Morosi
Copyright copy 2017 Mahmut Demirtas and Alkan Soysal This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited
In this paper we introduce nonoverlay microcellmacrocell planning that is optimally designed for improving energy efficiencyof the overall heterogeneous cellular network We consider two deployment strategies The first one is based on a fixed hexagonalgrid and the second one is based on a stochastic geometry In both of our models microcells are placed in those areas wherethe received signal power levels of macrocell common pilot channels are below a certain threshold Thus interference betweenmicrocells and macrocells is minimized As a result addition of microcells increases the achieved number of bits per unit energyUnder such deployment assumptions we investigate the effects of certain parameters on the energy efficiency These parametersinclude the user traffic the Intersite Distance (ISD) the size of microcells and the number of microcells per macrocell for the gridmodel and macrocell density and microcell density for the stochastic model The results of our performance analyses show thatutilizing microcells in a sparse user scenario is worse for the energy efficiency whereas it significantly improves both energy andspectral efficiencies in a dense user scenario Another interesting observation is that it is possible to choose an optimum numberof microcells for a given macrocell density
1 Introduction
For a long period of time the main concern of cellularsystemswas to increase the spectral efficiency Among severalothers one way to increase the spectral efficiency is to overlaymicrocells on the existing macrocell coverage area Thistwo-tier approach guarantees the coverage and increases thespectral efficiency of the users that are close to microcellstations However the downside of this approach is that theenergy efficiency of the network gets worse by the addition ofnew overlaid microcells
Over the last couple of years the energy efficiency ofcellular networks has been an increasing concern becauseof its environmental and operational cost effects In orderto improve the energy efficiency several solutions are pro-posed in the literature Detailed approach to general energyefficiency problem can be found in [1ndash5] and the referencestherein
The performance of an energy efficiency analysis dependsstrongly on the definition of energy efficiency metric [6]
Area power consumption and bit per joule are the two mostcommon energy efficiency metrics that are considered in theliterature Similar to [7ndash10] we considered area power con-sumption metric in some previous studies [11 12] Howeverother studies report that bit per joule metric captures theenergy efficiency in high-load conditions better than areapower consumption metric [2] In the literature there areseveral works where the authors employed bit per joule astheir efficiencymetric [13ndash17]We also use bit per joulemetricin this study since our focus is to obtain energy efficientmethods for increasingly high demands of spectral efficiency
Another important concept in the energy efficiency anal-ysis of heterogeneous cellular networks is the base stationdeployment model In the literature generally two differentmodels are used to determine the locations of base stationsfixed hexagonal grid model and Poisson Point Process (PPP)based stochastic geometrymodel Although stochastic geom-etry models are better fit to real base station deploymentsfixed hexagonal grid models can provide a better insight tothe mathematical problem
HindawiWireless Communications and Mobile ComputingVolume 2017 Article ID 6519709 11 pageshttpsdoiorg10115520176519709
2 Wireless Communications and Mobile Computing
In [7ndash12] fixed hexagonal grid model is employed andarea power consumption metric is used for energy efficiencycomparison The authors in [7 8] find the optimum ISDbetween macrocells when several microcells are overlaid onthe macrocell coverage area However in such a scenariomicrocell addition always increases the total power con-sumption of the system because of the overlay structure In[9] overlaid microcells or reduced range omnidirectionalmacrocells are turned on and off depending on a trafficmodel The authors find the optimum ISD for a range ofpath loss exponent values In [10] microcell base stationplanning is considered however the optimization problemis to minimize the number of microcell base stations over aset of candidate microcell base station positions and a trafficconstraint Therefore improving energy efficiency is not themain goal in [10] In [11 12] we found optimum ISD in anonoverlay microcell deployment
There are also some studies that consider bit per joulemetric for energy efficiency analysis In [13 15] the authorsconsider a single cell OFDMA network and investigate thetrade-off between spectral efficiency and bit per joule energyefficiency In [16 17] the authors extend their results toa case where there is a single macrocell base station withseveral uniformly overlaid small cells and to multiple inputmultiple output (MIMO) broadcast channels respectivelyIn a multicell scenario and for a fixed grid [14] assumes ahomogeneous network deploymentwithmicrocell or picocellbase stations and calculates the effect of backhaul powerconsumption on the bit per joule energy efficiency In [18] weconsidered a heterogeneous network with a fixed hexagonalgrid and showed that bit per joule energy efficiency increaseswith increasing number of microcells
Stochastic geometry based models are considered in [19ndash22] In [19] macrocells are located according to a PPP ina Euclidean plane The authors consider only homogeneousmacrocell networks and their goal is to find tractable cov-erage and rate expressions These mathematical analyses ofcoverage and average rates are extended to heterogeneousnetworks in [20] The model in [21] assumes a single macro-cell and several small cells that are distributed according toa PPP The authors find the optimal density of small cellsthat maximize energy efficiency Similar results for multicelloverlaid heterogeneous networks are derived in [22] whereenergy efficiency metric is area power consumption
In this paper we improve the energy efficiency through anonoverlay planning of heterogeneous networks We deploymicrocells at the locations where the received signal strengthis expected to be relatively low In the fixed grid model themicrocell locations are chosen to be the cell edges of thehexagonal cell site In the stochastic geometry model weemploy a two-stage deployment In the first stage macrocellsare placed according to a PPP and the coverage regions aredetermined In the second stage we detect the regions wherethe received signal strength is lower than a certain limitThen we place the microcells on those regions according to aseparate PPP and update the coverage regions
Due to the nonoverlay nature of our deployment a mac-rocell base station saves power when microcells are deployedin a site Using this model we calculate the energy efficiency
as a function of ISD (or macrocell density) the number ofmicrocells (or microcell density) and the size of microcellsFor power consumption modeling we use comprehensivepower consumption models that are introduced in [23]We consider bit per joule as our energy efficiency metricThrough our simulations we observe that deploying micro-cells simultaneously increases both energy efficiency andspectral efficiency Also we conclude that it is possible tochoose intervals for ISD (or macrocell density) and numberof microcells (or microcell density) that improves the energyefficiency the most
2 System Model
In this paper our goal is to improve the energy efficiencyof a nonoverlay heterogeneous cellular network withoutcompromising the spectral efficiency Here ldquoheterogeneouscellular networkrdquo refers to a single technology network thatcontains different sizes of base stations We assume a fixedcoverage constraint that guarantees that a certain minimumpercentage of a service area is covered In addition we assumethat all base stations work under full-load condition
In order to investigate the heterogeneous networks interms of energy and spectral efficiency we use two differentmodels for base station deployment a fixed hexagonal gridmodel and a stochastic geometry based model In the firstmodel we consider a hexagonal grid of macrocells whereeach macrocell receives interference from a tier of neighbor-ing macrocells Due to the nonoverlay nature of microcelldeployment and in order to save power the radius of amacro-cell might get smaller as the number of microcells increasesThis can be observed in Figures 1(a) 1(c) and 1(e) wherethe coverage constraint is 100 for all subfigures ISD deter-mines the hexagonal cell size and coverage area determinesthe macrocell radius 119877119898 (see Figure 1(a)) Microcells aredeployed along the edges of the hexagons (see Figures 1(c)and 1(e)) Our goal is to analyze the energy efficiency of sucha deployment over certain parameters like ISD number ofmicrocells and microcell radius We also consider differentuser densities in order to observe the effect of microcellutilization In the sparse scenario we have 5 userskm2whereas in the dense scenario we have 100 userskm2
In the second model we use two separate Poisson pointprocesses for macrocells and microcells with densities of 120582119872and 120582120583 respectively A 10 times 10 km2 area is chosen for theanalysis First macrocells are located with the density of 120582119872and the coverage regions are determined according to thereceived signal strength from the common pilot channel (seeFigure 1(b)) A point is said to be covered if the received signalstrength from the pilot channel of the closest base station isabove a certain limit namely 119875119888 The allocated power ratiofor the pilot channel is 10 of the total power budget of eachbase station If the pilot channel power is not enough to satisfythe coverage condition for a particular scenario we increasethe pilot signal strength by a step size of 2 of total powerbudget At the second stage the regions forwhich the receivedsignal strength is lower than another certain limit namely119875120583are detected Then microcells are deployed on those regionswith the density of 120582120583 and the coverage regions are updated
(see Figures 1(d) and 1(f)) It is important to note that 119875120583 isalways larger than119875119888Thus the regions that can be covered bymacrocells shrink as it is in the grid model By this methodwe keep nonoverlaying structure as much as possible
21 Power Consumption Models For an accurate energyefficiency analysis a power consumption model needs toinclude power consumed at the base stations due to signalprocessing cooling network transmission and so forth inaddition to the Radio Frequency (RF) transmitted powerThemore detailed power consumption modeling is done themore accurate our analysis will become
The RF transmit power is allocated mainly between thecommon pilot channel and the traffic channelsThe commonpilot channel received power determines the coverage areaand traffic channel received power determines the data ratethat is provided to the mobile stations In order to calculatethe received power at the mobile stations we considerdeterministic path loss shadowing and fast fading effectsThe received signal strength (in dB scale) is given as afunction of the distance between transmitter and receiver 119903as
119875rx (119903) = 119875tx minus PL (119903) minus 120585 minus 120577 (1)
where 119875tx is the transmitted signal strength PL(119903) is the pathloss 120585 is the log-normal shadowing variable and 120577 is therandom fast fading variable For our coverage calculationswe use urban macrocell and microcell path loss modelsshadowing variance fast fading margin carrier frequencyand antenna height values that are given in [24] In Table 1a complete summary of our propagation model is given
In this model the Line of Sight (LOS) probability for amacrocell Pr119872(LOS) is [24]
Pr119872 (LOS) = min 18119903 1 (1 minus 119890minus11990363) + 119890minus11990363 (2)
and the LOS probability for microcells Pr120583(LOS) is [24]
Pr120583 (LOS) = min 18119903 1 (1 minus 119890minus11990336) + 119890minus11990336 (3)
where 119903 is the distance between the user and serving basestation
As stated above RF transmit power is only a frac-tion of the total power consumed by a base station In
order to include the effects of baseband signal processingtransmission cooling and so forth we consider the powerconsumption model that is introduced in [23] as follows
where 119894 = 119872 and 119894 = 120583 correspond to macrocells and micro-cells respectively 119873TX119894 is the number of transceiver chains1198750119894 is the power consumption at zero RF output power Δ 119894is the slope of the load dependent power consumption and119875max119894 is the maximum power budget of base stations Herethe constant term 1198750119894 includes power consumed at the basestations due to signal processing cooling backhaul and soforth
22 Energy Efficiency Metric In literature area power con-sumption is frequently used as an efficiency metric forwireless cellular networks It is defined as the ratio of the totalpower consumption and the total service area of a networkThismetric has the advantage of capturing the size of the totalservice area However it cannot capture the effect of the totaldata rate that is provided in the same service area In orderto capture both effects in this work we consider the ratioof area spectral efficiency to area power consumption as ourperformance metric
EE = Area Spectral EfficiencyArea Power Consumption
(bitssHzkm2Wattskm2
= bitsjouleHz) (5)
Area power consumption is the per unit area total powerconsumed by all base stations and is calculated by summing(4) over all macrocell and microcell base stations in a givenservice area Area spectral efficiency was defined as the totaldata rate per unit area andper unit bandwidth that is providedby a base station in [25] Since there are many base stations ina service area we use the following formula [8] to evaluatethe total area spectral efficiency
ASE = sum119894
119878119894 Pr 119873119894 gt 0 (6)
where the summation is over all macrocell andmicrocell basestations 119878119894 is the area spectral efficiency of base station 119894 and
Wireless Communications and Mobile Computing 5
Table 2 Parameters for path loss models
Urban macrocell Urban microcell119875119888 (receiver sensitivity) minus120 dBm 119875119888 (receiver sensitivity) minus120 dBm119891119888 (carrier frequency in GHz) 24 GHz 119891119888 (carrier frequency in GHz) 24GHzℎBS (base station antenna height) 25 m ℎBS (base station antenna height) 10mℎUT (user terminal antenna height) 15m ℎUT (user terminal antenna height) 15mℎ1015840BS (effective BS antenna height) 24m ℎ1015840BS (effective BS antenna height) 9mℎ1015840UT (effective UT antenna height) 05m ℎ1015840UT (effective UT antenna height) 05m1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888m 1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888mℎ (average building height) 20m119882 (street width) 20m
119873119894 is the number of uniformly distributed users in the servicearea of base station 119894
The area spectral efficiency of a particular base station 119878119894is the per unit area total achievable data rate that is calculatedover the users in the service area of that base station and it isgiven as
where 120574119906(119903119906119894) is the Signal to Interference and Noise Ratio(SINR) of user 119906 119903119906119894 is the distance of user 119906 to base station 119894and 119860119862 is the total service area
Here we assume a universal frequency reuse In otherwords all signals from adjacent macrocells and microcells tobase station 119894 contribute to the interference level As a resultwe define the SINR level of user119906 that is served by base station119894 as
where 119875rx119906(119903119906119894) is the received power level from base station119894 to user 119906 and it can be calculated using (1)
3 Hexagonal Grid Model
In this section we calculate the energy efficiency perfor-mance of hexagonal grid deployment model over downlinkchannels using Monte Carlo simulations We generate a largenumber of uniformly distributed users in the service areaThen we calculate energy efficiency and spectral efficiencyof the network as a function of ISD number of microcellsuser density and microcell size We assume that the coverageconstraint is 95 We consider the macrocell ISD to be from500 m to 1500 m in order to reflect typical macrocell sizesThe parameters that we use in power consumption modelare given as the following 119873TX119872 = 6 1198750119872 = 130 WattsΔ119872 = 47 119875max119872 = 20 Watts 119873TX120583 = 2 1198750120583 = 56 WattsΔ 120583 = 26 and 119875max120583 = 63 Watts [23] In addition path lossparameters are given in Table 2 [24]
In Figure 2 we plot the energy efficiency of severalmicrocell density scenarios with respect to ISD We assumethat the size of a microcell is 20 of the size of a macrocell
We start with the sparse scenario with user density of 5userskm2 First observation of Figure 2(a) is that the energyefficiency improves with increasing ISD except for the homo-geneous scenario However deploying microcells results ina worse energy efficiency under sparse user assumptionand for low ISD values This is basically a result of manymicrocells operating under no load condition but spendingconsiderable amount of power just to stay on For lowISD values the power of those microcells with no userscontributes negatively (in the form of interference) to thespectral efficiency of neighbor cells but does not providespectral efficiency for its own cellTherefore we conclude thatutilizing microcells for low traffic areas is not necessary in afixed hexagonal deployment with small ISD
Next we analyze the performance of a dense user sce-nario with 100 userskm2 We observe in Figure 2(b) thatsimilar to sparse user case the energy efficiency improveswith increasing ISD However for a given ISD the energyefficiency is much better in dense user case than it is in sparseuser case The reason for this is that most microcells are fullyutilized under dense user scenarios Another observation isthat the energy efficiency is not monotonic in the number ofmicrocells After some point adding more microcells doesnot further improve the energy efficiency Therefore theremust be an optimum number of microcells as a function ofuser traffic in a given area For the scenario in Figure 2(b)8microcellssite is almost 5 times more energy efficient thanmacrocell-only network
In a dense user scenario increasing the number of nono-verlay microcells improves the energy efficiency becausemacrocells shrink their coverage area in order to save sometransmission power and do not spent power to service celledge users However we should also investigate how muchthis approach affects the spectral efficiency We first considerthe sparse scenario with 5 userskm2 In Figure 3(a) weobserve that the spectral efficiency decreases with increasingISD The reason for this is the decrease in expected receivedsignal strength with the increase in cell size For the puremacrocell case the decrease in spectral efficiency is muchfaster We also observe that 8 microcellssite is the mostspectral efficient scenario In this analysis we do not includehigher microcell densities than 11 microcellssite becausethe microcells start to overlap considerably when we deploymore than 11 microcells per site Next we consider a dense
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
In [7ndash12] fixed hexagonal grid model is employed andarea power consumption metric is used for energy efficiencycomparison The authors in [7 8] find the optimum ISDbetween macrocells when several microcells are overlaid onthe macrocell coverage area However in such a scenariomicrocell addition always increases the total power con-sumption of the system because of the overlay structure In[9] overlaid microcells or reduced range omnidirectionalmacrocells are turned on and off depending on a trafficmodel The authors find the optimum ISD for a range ofpath loss exponent values In [10] microcell base stationplanning is considered however the optimization problemis to minimize the number of microcell base stations over aset of candidate microcell base station positions and a trafficconstraint Therefore improving energy efficiency is not themain goal in [10] In [11 12] we found optimum ISD in anonoverlay microcell deployment
There are also some studies that consider bit per joulemetric for energy efficiency analysis In [13 15] the authorsconsider a single cell OFDMA network and investigate thetrade-off between spectral efficiency and bit per joule energyefficiency In [16 17] the authors extend their results toa case where there is a single macrocell base station withseveral uniformly overlaid small cells and to multiple inputmultiple output (MIMO) broadcast channels respectivelyIn a multicell scenario and for a fixed grid [14] assumes ahomogeneous network deploymentwithmicrocell or picocellbase stations and calculates the effect of backhaul powerconsumption on the bit per joule energy efficiency In [18] weconsidered a heterogeneous network with a fixed hexagonalgrid and showed that bit per joule energy efficiency increaseswith increasing number of microcells
Stochastic geometry based models are considered in [19ndash22] In [19] macrocells are located according to a PPP ina Euclidean plane The authors consider only homogeneousmacrocell networks and their goal is to find tractable cov-erage and rate expressions These mathematical analyses ofcoverage and average rates are extended to heterogeneousnetworks in [20] The model in [21] assumes a single macro-cell and several small cells that are distributed according toa PPP The authors find the optimal density of small cellsthat maximize energy efficiency Similar results for multicelloverlaid heterogeneous networks are derived in [22] whereenergy efficiency metric is area power consumption
In this paper we improve the energy efficiency through anonoverlay planning of heterogeneous networks We deploymicrocells at the locations where the received signal strengthis expected to be relatively low In the fixed grid model themicrocell locations are chosen to be the cell edges of thehexagonal cell site In the stochastic geometry model weemploy a two-stage deployment In the first stage macrocellsare placed according to a PPP and the coverage regions aredetermined In the second stage we detect the regions wherethe received signal strength is lower than a certain limitThen we place the microcells on those regions according to aseparate PPP and update the coverage regions
Due to the nonoverlay nature of our deployment a mac-rocell base station saves power when microcells are deployedin a site Using this model we calculate the energy efficiency
as a function of ISD (or macrocell density) the number ofmicrocells (or microcell density) and the size of microcellsFor power consumption modeling we use comprehensivepower consumption models that are introduced in [23]We consider bit per joule as our energy efficiency metricThrough our simulations we observe that deploying micro-cells simultaneously increases both energy efficiency andspectral efficiency Also we conclude that it is possible tochoose intervals for ISD (or macrocell density) and numberof microcells (or microcell density) that improves the energyefficiency the most
2 System Model
In this paper our goal is to improve the energy efficiencyof a nonoverlay heterogeneous cellular network withoutcompromising the spectral efficiency Here ldquoheterogeneouscellular networkrdquo refers to a single technology network thatcontains different sizes of base stations We assume a fixedcoverage constraint that guarantees that a certain minimumpercentage of a service area is covered In addition we assumethat all base stations work under full-load condition
In order to investigate the heterogeneous networks interms of energy and spectral efficiency we use two differentmodels for base station deployment a fixed hexagonal gridmodel and a stochastic geometry based model In the firstmodel we consider a hexagonal grid of macrocells whereeach macrocell receives interference from a tier of neighbor-ing macrocells Due to the nonoverlay nature of microcelldeployment and in order to save power the radius of amacro-cell might get smaller as the number of microcells increasesThis can be observed in Figures 1(a) 1(c) and 1(e) wherethe coverage constraint is 100 for all subfigures ISD deter-mines the hexagonal cell size and coverage area determinesthe macrocell radius 119877119898 (see Figure 1(a)) Microcells aredeployed along the edges of the hexagons (see Figures 1(c)and 1(e)) Our goal is to analyze the energy efficiency of sucha deployment over certain parameters like ISD number ofmicrocells and microcell radius We also consider differentuser densities in order to observe the effect of microcellutilization In the sparse scenario we have 5 userskm2whereas in the dense scenario we have 100 userskm2
In the second model we use two separate Poisson pointprocesses for macrocells and microcells with densities of 120582119872and 120582120583 respectively A 10 times 10 km2 area is chosen for theanalysis First macrocells are located with the density of 120582119872and the coverage regions are determined according to thereceived signal strength from the common pilot channel (seeFigure 1(b)) A point is said to be covered if the received signalstrength from the pilot channel of the closest base station isabove a certain limit namely 119875119888 The allocated power ratiofor the pilot channel is 10 of the total power budget of eachbase station If the pilot channel power is not enough to satisfythe coverage condition for a particular scenario we increasethe pilot signal strength by a step size of 2 of total powerbudget At the second stage the regions forwhich the receivedsignal strength is lower than another certain limit namely119875120583are detected Then microcells are deployed on those regionswith the density of 120582120583 and the coverage regions are updated
(see Figures 1(d) and 1(f)) It is important to note that 119875120583 isalways larger than119875119888Thus the regions that can be covered bymacrocells shrink as it is in the grid model By this methodwe keep nonoverlaying structure as much as possible
21 Power Consumption Models For an accurate energyefficiency analysis a power consumption model needs toinclude power consumed at the base stations due to signalprocessing cooling network transmission and so forth inaddition to the Radio Frequency (RF) transmitted powerThemore detailed power consumption modeling is done themore accurate our analysis will become
The RF transmit power is allocated mainly between thecommon pilot channel and the traffic channelsThe commonpilot channel received power determines the coverage areaand traffic channel received power determines the data ratethat is provided to the mobile stations In order to calculatethe received power at the mobile stations we considerdeterministic path loss shadowing and fast fading effectsThe received signal strength (in dB scale) is given as afunction of the distance between transmitter and receiver 119903as
119875rx (119903) = 119875tx minus PL (119903) minus 120585 minus 120577 (1)
where 119875tx is the transmitted signal strength PL(119903) is the pathloss 120585 is the log-normal shadowing variable and 120577 is therandom fast fading variable For our coverage calculationswe use urban macrocell and microcell path loss modelsshadowing variance fast fading margin carrier frequencyand antenna height values that are given in [24] In Table 1a complete summary of our propagation model is given
In this model the Line of Sight (LOS) probability for amacrocell Pr119872(LOS) is [24]
Pr119872 (LOS) = min 18119903 1 (1 minus 119890minus11990363) + 119890minus11990363 (2)
and the LOS probability for microcells Pr120583(LOS) is [24]
Pr120583 (LOS) = min 18119903 1 (1 minus 119890minus11990336) + 119890minus11990336 (3)
where 119903 is the distance between the user and serving basestation
As stated above RF transmit power is only a frac-tion of the total power consumed by a base station In
order to include the effects of baseband signal processingtransmission cooling and so forth we consider the powerconsumption model that is introduced in [23] as follows
where 119894 = 119872 and 119894 = 120583 correspond to macrocells and micro-cells respectively 119873TX119894 is the number of transceiver chains1198750119894 is the power consumption at zero RF output power Δ 119894is the slope of the load dependent power consumption and119875max119894 is the maximum power budget of base stations Herethe constant term 1198750119894 includes power consumed at the basestations due to signal processing cooling backhaul and soforth
22 Energy Efficiency Metric In literature area power con-sumption is frequently used as an efficiency metric forwireless cellular networks It is defined as the ratio of the totalpower consumption and the total service area of a networkThismetric has the advantage of capturing the size of the totalservice area However it cannot capture the effect of the totaldata rate that is provided in the same service area In orderto capture both effects in this work we consider the ratioof area spectral efficiency to area power consumption as ourperformance metric
EE = Area Spectral EfficiencyArea Power Consumption
(bitssHzkm2Wattskm2
= bitsjouleHz) (5)
Area power consumption is the per unit area total powerconsumed by all base stations and is calculated by summing(4) over all macrocell and microcell base stations in a givenservice area Area spectral efficiency was defined as the totaldata rate per unit area andper unit bandwidth that is providedby a base station in [25] Since there are many base stations ina service area we use the following formula [8] to evaluatethe total area spectral efficiency
ASE = sum119894
119878119894 Pr 119873119894 gt 0 (6)
where the summation is over all macrocell andmicrocell basestations 119878119894 is the area spectral efficiency of base station 119894 and
Wireless Communications and Mobile Computing 5
Table 2 Parameters for path loss models
Urban macrocell Urban microcell119875119888 (receiver sensitivity) minus120 dBm 119875119888 (receiver sensitivity) minus120 dBm119891119888 (carrier frequency in GHz) 24 GHz 119891119888 (carrier frequency in GHz) 24GHzℎBS (base station antenna height) 25 m ℎBS (base station antenna height) 10mℎUT (user terminal antenna height) 15m ℎUT (user terminal antenna height) 15mℎ1015840BS (effective BS antenna height) 24m ℎ1015840BS (effective BS antenna height) 9mℎ1015840UT (effective UT antenna height) 05m ℎ1015840UT (effective UT antenna height) 05m1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888m 1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888mℎ (average building height) 20m119882 (street width) 20m
119873119894 is the number of uniformly distributed users in the servicearea of base station 119894
The area spectral efficiency of a particular base station 119878119894is the per unit area total achievable data rate that is calculatedover the users in the service area of that base station and it isgiven as
where 120574119906(119903119906119894) is the Signal to Interference and Noise Ratio(SINR) of user 119906 119903119906119894 is the distance of user 119906 to base station 119894and 119860119862 is the total service area
Here we assume a universal frequency reuse In otherwords all signals from adjacent macrocells and microcells tobase station 119894 contribute to the interference level As a resultwe define the SINR level of user119906 that is served by base station119894 as
where 119875rx119906(119903119906119894) is the received power level from base station119894 to user 119906 and it can be calculated using (1)
3 Hexagonal Grid Model
In this section we calculate the energy efficiency perfor-mance of hexagonal grid deployment model over downlinkchannels using Monte Carlo simulations We generate a largenumber of uniformly distributed users in the service areaThen we calculate energy efficiency and spectral efficiencyof the network as a function of ISD number of microcellsuser density and microcell size We assume that the coverageconstraint is 95 We consider the macrocell ISD to be from500 m to 1500 m in order to reflect typical macrocell sizesThe parameters that we use in power consumption modelare given as the following 119873TX119872 = 6 1198750119872 = 130 WattsΔ119872 = 47 119875max119872 = 20 Watts 119873TX120583 = 2 1198750120583 = 56 WattsΔ 120583 = 26 and 119875max120583 = 63 Watts [23] In addition path lossparameters are given in Table 2 [24]
In Figure 2 we plot the energy efficiency of severalmicrocell density scenarios with respect to ISD We assumethat the size of a microcell is 20 of the size of a macrocell
We start with the sparse scenario with user density of 5userskm2 First observation of Figure 2(a) is that the energyefficiency improves with increasing ISD except for the homo-geneous scenario However deploying microcells results ina worse energy efficiency under sparse user assumptionand for low ISD values This is basically a result of manymicrocells operating under no load condition but spendingconsiderable amount of power just to stay on For lowISD values the power of those microcells with no userscontributes negatively (in the form of interference) to thespectral efficiency of neighbor cells but does not providespectral efficiency for its own cellTherefore we conclude thatutilizing microcells for low traffic areas is not necessary in afixed hexagonal deployment with small ISD
Next we analyze the performance of a dense user sce-nario with 100 userskm2 We observe in Figure 2(b) thatsimilar to sparse user case the energy efficiency improveswith increasing ISD However for a given ISD the energyefficiency is much better in dense user case than it is in sparseuser case The reason for this is that most microcells are fullyutilized under dense user scenarios Another observation isthat the energy efficiency is not monotonic in the number ofmicrocells After some point adding more microcells doesnot further improve the energy efficiency Therefore theremust be an optimum number of microcells as a function ofuser traffic in a given area For the scenario in Figure 2(b)8microcellssite is almost 5 times more energy efficient thanmacrocell-only network
In a dense user scenario increasing the number of nono-verlay microcells improves the energy efficiency becausemacrocells shrink their coverage area in order to save sometransmission power and do not spent power to service celledge users However we should also investigate how muchthis approach affects the spectral efficiency We first considerthe sparse scenario with 5 userskm2 In Figure 3(a) weobserve that the spectral efficiency decreases with increasingISD The reason for this is the decrease in expected receivedsignal strength with the increase in cell size For the puremacrocell case the decrease in spectral efficiency is muchfaster We also observe that 8 microcellssite is the mostspectral efficient scenario In this analysis we do not includehigher microcell densities than 11 microcellssite becausethe microcells start to overlap considerably when we deploymore than 11 microcells per site Next we consider a dense
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
(see Figures 1(d) and 1(f)) It is important to note that 119875120583 isalways larger than119875119888Thus the regions that can be covered bymacrocells shrink as it is in the grid model By this methodwe keep nonoverlaying structure as much as possible
21 Power Consumption Models For an accurate energyefficiency analysis a power consumption model needs toinclude power consumed at the base stations due to signalprocessing cooling network transmission and so forth inaddition to the Radio Frequency (RF) transmitted powerThemore detailed power consumption modeling is done themore accurate our analysis will become
The RF transmit power is allocated mainly between thecommon pilot channel and the traffic channelsThe commonpilot channel received power determines the coverage areaand traffic channel received power determines the data ratethat is provided to the mobile stations In order to calculatethe received power at the mobile stations we considerdeterministic path loss shadowing and fast fading effectsThe received signal strength (in dB scale) is given as afunction of the distance between transmitter and receiver 119903as
119875rx (119903) = 119875tx minus PL (119903) minus 120585 minus 120577 (1)
where 119875tx is the transmitted signal strength PL(119903) is the pathloss 120585 is the log-normal shadowing variable and 120577 is therandom fast fading variable For our coverage calculationswe use urban macrocell and microcell path loss modelsshadowing variance fast fading margin carrier frequencyand antenna height values that are given in [24] In Table 1a complete summary of our propagation model is given
In this model the Line of Sight (LOS) probability for amacrocell Pr119872(LOS) is [24]
Pr119872 (LOS) = min 18119903 1 (1 minus 119890minus11990363) + 119890minus11990363 (2)
and the LOS probability for microcells Pr120583(LOS) is [24]
Pr120583 (LOS) = min 18119903 1 (1 minus 119890minus11990336) + 119890minus11990336 (3)
where 119903 is the distance between the user and serving basestation
As stated above RF transmit power is only a frac-tion of the total power consumed by a base station In
order to include the effects of baseband signal processingtransmission cooling and so forth we consider the powerconsumption model that is introduced in [23] as follows
where 119894 = 119872 and 119894 = 120583 correspond to macrocells and micro-cells respectively 119873TX119894 is the number of transceiver chains1198750119894 is the power consumption at zero RF output power Δ 119894is the slope of the load dependent power consumption and119875max119894 is the maximum power budget of base stations Herethe constant term 1198750119894 includes power consumed at the basestations due to signal processing cooling backhaul and soforth
22 Energy Efficiency Metric In literature area power con-sumption is frequently used as an efficiency metric forwireless cellular networks It is defined as the ratio of the totalpower consumption and the total service area of a networkThismetric has the advantage of capturing the size of the totalservice area However it cannot capture the effect of the totaldata rate that is provided in the same service area In orderto capture both effects in this work we consider the ratioof area spectral efficiency to area power consumption as ourperformance metric
EE = Area Spectral EfficiencyArea Power Consumption
(bitssHzkm2Wattskm2
= bitsjouleHz) (5)
Area power consumption is the per unit area total powerconsumed by all base stations and is calculated by summing(4) over all macrocell and microcell base stations in a givenservice area Area spectral efficiency was defined as the totaldata rate per unit area andper unit bandwidth that is providedby a base station in [25] Since there are many base stations ina service area we use the following formula [8] to evaluatethe total area spectral efficiency
ASE = sum119894
119878119894 Pr 119873119894 gt 0 (6)
where the summation is over all macrocell andmicrocell basestations 119878119894 is the area spectral efficiency of base station 119894 and
Wireless Communications and Mobile Computing 5
Table 2 Parameters for path loss models
Urban macrocell Urban microcell119875119888 (receiver sensitivity) minus120 dBm 119875119888 (receiver sensitivity) minus120 dBm119891119888 (carrier frequency in GHz) 24 GHz 119891119888 (carrier frequency in GHz) 24GHzℎBS (base station antenna height) 25 m ℎBS (base station antenna height) 10mℎUT (user terminal antenna height) 15m ℎUT (user terminal antenna height) 15mℎ1015840BS (effective BS antenna height) 24m ℎ1015840BS (effective BS antenna height) 9mℎ1015840UT (effective UT antenna height) 05m ℎ1015840UT (effective UT antenna height) 05m1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888m 1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888mℎ (average building height) 20m119882 (street width) 20m
119873119894 is the number of uniformly distributed users in the servicearea of base station 119894
The area spectral efficiency of a particular base station 119878119894is the per unit area total achievable data rate that is calculatedover the users in the service area of that base station and it isgiven as
where 120574119906(119903119906119894) is the Signal to Interference and Noise Ratio(SINR) of user 119906 119903119906119894 is the distance of user 119906 to base station 119894and 119860119862 is the total service area
Here we assume a universal frequency reuse In otherwords all signals from adjacent macrocells and microcells tobase station 119894 contribute to the interference level As a resultwe define the SINR level of user119906 that is served by base station119894 as
where 119875rx119906(119903119906119894) is the received power level from base station119894 to user 119906 and it can be calculated using (1)
3 Hexagonal Grid Model
In this section we calculate the energy efficiency perfor-mance of hexagonal grid deployment model over downlinkchannels using Monte Carlo simulations We generate a largenumber of uniformly distributed users in the service areaThen we calculate energy efficiency and spectral efficiencyof the network as a function of ISD number of microcellsuser density and microcell size We assume that the coverageconstraint is 95 We consider the macrocell ISD to be from500 m to 1500 m in order to reflect typical macrocell sizesThe parameters that we use in power consumption modelare given as the following 119873TX119872 = 6 1198750119872 = 130 WattsΔ119872 = 47 119875max119872 = 20 Watts 119873TX120583 = 2 1198750120583 = 56 WattsΔ 120583 = 26 and 119875max120583 = 63 Watts [23] In addition path lossparameters are given in Table 2 [24]
In Figure 2 we plot the energy efficiency of severalmicrocell density scenarios with respect to ISD We assumethat the size of a microcell is 20 of the size of a macrocell
We start with the sparse scenario with user density of 5userskm2 First observation of Figure 2(a) is that the energyefficiency improves with increasing ISD except for the homo-geneous scenario However deploying microcells results ina worse energy efficiency under sparse user assumptionand for low ISD values This is basically a result of manymicrocells operating under no load condition but spendingconsiderable amount of power just to stay on For lowISD values the power of those microcells with no userscontributes negatively (in the form of interference) to thespectral efficiency of neighbor cells but does not providespectral efficiency for its own cellTherefore we conclude thatutilizing microcells for low traffic areas is not necessary in afixed hexagonal deployment with small ISD
Next we analyze the performance of a dense user sce-nario with 100 userskm2 We observe in Figure 2(b) thatsimilar to sparse user case the energy efficiency improveswith increasing ISD However for a given ISD the energyefficiency is much better in dense user case than it is in sparseuser case The reason for this is that most microcells are fullyutilized under dense user scenarios Another observation isthat the energy efficiency is not monotonic in the number ofmicrocells After some point adding more microcells doesnot further improve the energy efficiency Therefore theremust be an optimum number of microcells as a function ofuser traffic in a given area For the scenario in Figure 2(b)8microcellssite is almost 5 times more energy efficient thanmacrocell-only network
In a dense user scenario increasing the number of nono-verlay microcells improves the energy efficiency becausemacrocells shrink their coverage area in order to save sometransmission power and do not spent power to service celledge users However we should also investigate how muchthis approach affects the spectral efficiency We first considerthe sparse scenario with 5 userskm2 In Figure 3(a) weobserve that the spectral efficiency decreases with increasingISD The reason for this is the decrease in expected receivedsignal strength with the increase in cell size For the puremacrocell case the decrease in spectral efficiency is muchfaster We also observe that 8 microcellssite is the mostspectral efficient scenario In this analysis we do not includehigher microcell densities than 11 microcellssite becausethe microcells start to overlap considerably when we deploymore than 11 microcells per site Next we consider a dense
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
(see Figures 1(d) and 1(f)) It is important to note that 119875120583 isalways larger than119875119888Thus the regions that can be covered bymacrocells shrink as it is in the grid model By this methodwe keep nonoverlaying structure as much as possible
21 Power Consumption Models For an accurate energyefficiency analysis a power consumption model needs toinclude power consumed at the base stations due to signalprocessing cooling network transmission and so forth inaddition to the Radio Frequency (RF) transmitted powerThemore detailed power consumption modeling is done themore accurate our analysis will become
The RF transmit power is allocated mainly between thecommon pilot channel and the traffic channelsThe commonpilot channel received power determines the coverage areaand traffic channel received power determines the data ratethat is provided to the mobile stations In order to calculatethe received power at the mobile stations we considerdeterministic path loss shadowing and fast fading effectsThe received signal strength (in dB scale) is given as afunction of the distance between transmitter and receiver 119903as
119875rx (119903) = 119875tx minus PL (119903) minus 120585 minus 120577 (1)
where 119875tx is the transmitted signal strength PL(119903) is the pathloss 120585 is the log-normal shadowing variable and 120577 is therandom fast fading variable For our coverage calculationswe use urban macrocell and microcell path loss modelsshadowing variance fast fading margin carrier frequencyand antenna height values that are given in [24] In Table 1a complete summary of our propagation model is given
In this model the Line of Sight (LOS) probability for amacrocell Pr119872(LOS) is [24]
Pr119872 (LOS) = min 18119903 1 (1 minus 119890minus11990363) + 119890minus11990363 (2)
and the LOS probability for microcells Pr120583(LOS) is [24]
Pr120583 (LOS) = min 18119903 1 (1 minus 119890minus11990336) + 119890minus11990336 (3)
where 119903 is the distance between the user and serving basestation
As stated above RF transmit power is only a frac-tion of the total power consumed by a base station In
order to include the effects of baseband signal processingtransmission cooling and so forth we consider the powerconsumption model that is introduced in [23] as follows
where 119894 = 119872 and 119894 = 120583 correspond to macrocells and micro-cells respectively 119873TX119894 is the number of transceiver chains1198750119894 is the power consumption at zero RF output power Δ 119894is the slope of the load dependent power consumption and119875max119894 is the maximum power budget of base stations Herethe constant term 1198750119894 includes power consumed at the basestations due to signal processing cooling backhaul and soforth
22 Energy Efficiency Metric In literature area power con-sumption is frequently used as an efficiency metric forwireless cellular networks It is defined as the ratio of the totalpower consumption and the total service area of a networkThismetric has the advantage of capturing the size of the totalservice area However it cannot capture the effect of the totaldata rate that is provided in the same service area In orderto capture both effects in this work we consider the ratioof area spectral efficiency to area power consumption as ourperformance metric
EE = Area Spectral EfficiencyArea Power Consumption
(bitssHzkm2Wattskm2
= bitsjouleHz) (5)
Area power consumption is the per unit area total powerconsumed by all base stations and is calculated by summing(4) over all macrocell and microcell base stations in a givenservice area Area spectral efficiency was defined as the totaldata rate per unit area andper unit bandwidth that is providedby a base station in [25] Since there are many base stations ina service area we use the following formula [8] to evaluatethe total area spectral efficiency
ASE = sum119894
119878119894 Pr 119873119894 gt 0 (6)
where the summation is over all macrocell andmicrocell basestations 119878119894 is the area spectral efficiency of base station 119894 and
Wireless Communications and Mobile Computing 5
Table 2 Parameters for path loss models
Urban macrocell Urban microcell119875119888 (receiver sensitivity) minus120 dBm 119875119888 (receiver sensitivity) minus120 dBm119891119888 (carrier frequency in GHz) 24 GHz 119891119888 (carrier frequency in GHz) 24GHzℎBS (base station antenna height) 25 m ℎBS (base station antenna height) 10mℎUT (user terminal antenna height) 15m ℎUT (user terminal antenna height) 15mℎ1015840BS (effective BS antenna height) 24m ℎ1015840BS (effective BS antenna height) 9mℎ1015840UT (effective UT antenna height) 05m ℎ1015840UT (effective UT antenna height) 05m1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888m 1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888mℎ (average building height) 20m119882 (street width) 20m
119873119894 is the number of uniformly distributed users in the servicearea of base station 119894
The area spectral efficiency of a particular base station 119878119894is the per unit area total achievable data rate that is calculatedover the users in the service area of that base station and it isgiven as
where 120574119906(119903119906119894) is the Signal to Interference and Noise Ratio(SINR) of user 119906 119903119906119894 is the distance of user 119906 to base station 119894and 119860119862 is the total service area
Here we assume a universal frequency reuse In otherwords all signals from adjacent macrocells and microcells tobase station 119894 contribute to the interference level As a resultwe define the SINR level of user119906 that is served by base station119894 as
where 119875rx119906(119903119906119894) is the received power level from base station119894 to user 119906 and it can be calculated using (1)
3 Hexagonal Grid Model
In this section we calculate the energy efficiency perfor-mance of hexagonal grid deployment model over downlinkchannels using Monte Carlo simulations We generate a largenumber of uniformly distributed users in the service areaThen we calculate energy efficiency and spectral efficiencyof the network as a function of ISD number of microcellsuser density and microcell size We assume that the coverageconstraint is 95 We consider the macrocell ISD to be from500 m to 1500 m in order to reflect typical macrocell sizesThe parameters that we use in power consumption modelare given as the following 119873TX119872 = 6 1198750119872 = 130 WattsΔ119872 = 47 119875max119872 = 20 Watts 119873TX120583 = 2 1198750120583 = 56 WattsΔ 120583 = 26 and 119875max120583 = 63 Watts [23] In addition path lossparameters are given in Table 2 [24]
In Figure 2 we plot the energy efficiency of severalmicrocell density scenarios with respect to ISD We assumethat the size of a microcell is 20 of the size of a macrocell
We start with the sparse scenario with user density of 5userskm2 First observation of Figure 2(a) is that the energyefficiency improves with increasing ISD except for the homo-geneous scenario However deploying microcells results ina worse energy efficiency under sparse user assumptionand for low ISD values This is basically a result of manymicrocells operating under no load condition but spendingconsiderable amount of power just to stay on For lowISD values the power of those microcells with no userscontributes negatively (in the form of interference) to thespectral efficiency of neighbor cells but does not providespectral efficiency for its own cellTherefore we conclude thatutilizing microcells for low traffic areas is not necessary in afixed hexagonal deployment with small ISD
Next we analyze the performance of a dense user sce-nario with 100 userskm2 We observe in Figure 2(b) thatsimilar to sparse user case the energy efficiency improveswith increasing ISD However for a given ISD the energyefficiency is much better in dense user case than it is in sparseuser case The reason for this is that most microcells are fullyutilized under dense user scenarios Another observation isthat the energy efficiency is not monotonic in the number ofmicrocells After some point adding more microcells doesnot further improve the energy efficiency Therefore theremust be an optimum number of microcells as a function ofuser traffic in a given area For the scenario in Figure 2(b)8microcellssite is almost 5 times more energy efficient thanmacrocell-only network
In a dense user scenario increasing the number of nono-verlay microcells improves the energy efficiency becausemacrocells shrink their coverage area in order to save sometransmission power and do not spent power to service celledge users However we should also investigate how muchthis approach affects the spectral efficiency We first considerthe sparse scenario with 5 userskm2 In Figure 3(a) weobserve that the spectral efficiency decreases with increasingISD The reason for this is the decrease in expected receivedsignal strength with the increase in cell size For the puremacrocell case the decrease in spectral efficiency is muchfaster We also observe that 8 microcellssite is the mostspectral efficient scenario In this analysis we do not includehigher microcell densities than 11 microcellssite becausethe microcells start to overlap considerably when we deploymore than 11 microcells per site Next we consider a dense
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Urban macrocell Urban microcell119875119888 (receiver sensitivity) minus120 dBm 119875119888 (receiver sensitivity) minus120 dBm119891119888 (carrier frequency in GHz) 24 GHz 119891119888 (carrier frequency in GHz) 24GHzℎBS (base station antenna height) 25 m ℎBS (base station antenna height) 10mℎUT (user terminal antenna height) 15m ℎUT (user terminal antenna height) 15mℎ1015840BS (effective BS antenna height) 24m ℎ1015840BS (effective BS antenna height) 9mℎ1015840UT (effective UT antenna height) 05m ℎ1015840UT (effective UT antenna height) 05m1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888m 1199031015840BP (break point distance) 4ℎ1015840BSℎ1015840UT119891119888119888mℎ (average building height) 20m119882 (street width) 20m
119873119894 is the number of uniformly distributed users in the servicearea of base station 119894
The area spectral efficiency of a particular base station 119878119894is the per unit area total achievable data rate that is calculatedover the users in the service area of that base station and it isgiven as
where 120574119906(119903119906119894) is the Signal to Interference and Noise Ratio(SINR) of user 119906 119903119906119894 is the distance of user 119906 to base station 119894and 119860119862 is the total service area
Here we assume a universal frequency reuse In otherwords all signals from adjacent macrocells and microcells tobase station 119894 contribute to the interference level As a resultwe define the SINR level of user119906 that is served by base station119894 as
where 119875rx119906(119903119906119894) is the received power level from base station119894 to user 119906 and it can be calculated using (1)
3 Hexagonal Grid Model
In this section we calculate the energy efficiency perfor-mance of hexagonal grid deployment model over downlinkchannels using Monte Carlo simulations We generate a largenumber of uniformly distributed users in the service areaThen we calculate energy efficiency and spectral efficiencyof the network as a function of ISD number of microcellsuser density and microcell size We assume that the coverageconstraint is 95 We consider the macrocell ISD to be from500 m to 1500 m in order to reflect typical macrocell sizesThe parameters that we use in power consumption modelare given as the following 119873TX119872 = 6 1198750119872 = 130 WattsΔ119872 = 47 119875max119872 = 20 Watts 119873TX120583 = 2 1198750120583 = 56 WattsΔ 120583 = 26 and 119875max120583 = 63 Watts [23] In addition path lossparameters are given in Table 2 [24]
In Figure 2 we plot the energy efficiency of severalmicrocell density scenarios with respect to ISD We assumethat the size of a microcell is 20 of the size of a macrocell
We start with the sparse scenario with user density of 5userskm2 First observation of Figure 2(a) is that the energyefficiency improves with increasing ISD except for the homo-geneous scenario However deploying microcells results ina worse energy efficiency under sparse user assumptionand for low ISD values This is basically a result of manymicrocells operating under no load condition but spendingconsiderable amount of power just to stay on For lowISD values the power of those microcells with no userscontributes negatively (in the form of interference) to thespectral efficiency of neighbor cells but does not providespectral efficiency for its own cellTherefore we conclude thatutilizing microcells for low traffic areas is not necessary in afixed hexagonal deployment with small ISD
Next we analyze the performance of a dense user sce-nario with 100 userskm2 We observe in Figure 2(b) thatsimilar to sparse user case the energy efficiency improveswith increasing ISD However for a given ISD the energyefficiency is much better in dense user case than it is in sparseuser case The reason for this is that most microcells are fullyutilized under dense user scenarios Another observation isthat the energy efficiency is not monotonic in the number ofmicrocells After some point adding more microcells doesnot further improve the energy efficiency Therefore theremust be an optimum number of microcells as a function ofuser traffic in a given area For the scenario in Figure 2(b)8microcellssite is almost 5 times more energy efficient thanmacrocell-only network
In a dense user scenario increasing the number of nono-verlay microcells improves the energy efficiency becausemacrocells shrink their coverage area in order to save sometransmission power and do not spent power to service celledge users However we should also investigate how muchthis approach affects the spectral efficiency We first considerthe sparse scenario with 5 userskm2 In Figure 3(a) weobserve that the spectral efficiency decreases with increasingISD The reason for this is the decrease in expected receivedsignal strength with the increase in cell size For the puremacrocell case the decrease in spectral efficiency is muchfaster We also observe that 8 microcellssite is the mostspectral efficient scenario In this analysis we do not includehigher microcell densities than 11 microcellssite becausethe microcells start to overlap considerably when we deploymore than 11 microcells per site Next we consider a dense
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Pure macrocell1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 3 Area spectral efficiency for 20 microcell size
scenario with 100 userskm2 In Figure 3(b) we again observethat the spectral efficiency decreases with increasing ISDHowever it is important to note that the area spectralefficiency significantly increases with the increasing userdensity
Spectral efficiency results confirm that utilizing nonover-lay microcells is beneficial under dense user scenarios
Combining this with the energy efficiency performancewe conclude that utilizing nonoverlay microcells improvesboth spectral efficiency and energy efficiency under highuser traffic conditions On the other hand increasing ISDimproves energy efficiencywhile it reduces spectral efficiencyIn the remaining analysis we fix the ISD at 500 m to achievethe highest possible spectral efficiency
Wireless Communications and Mobile Computing 7
0
0002
0004
0006
0008
001
0012
0014
0016
0018
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
(a) Energy efficiency
20
40
60
80
100
120
140
160
180
2 4 6 8 10 120Number of microcells
5 userskm2 ISD 500100 userskm2 ISD 500
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Figure 4 The effect of the number of microcells when ISD = 500 m and microcell size is 20
0004
0006
0008
001
0012
0014
0016
Ener
gy effi
cien
cy (b
itH
zJ)
015 02 025 03 035 04 045 0501Microcell radius ratio
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
(a) Energy efficiency
015 02 025 03 035 04 045 0501Microcell radius ratio
20
40
60
80
100
120
140
160
180
200
220
240
1 microcell per site2 microcells per site35 microcells per site
5 microcells per site8 microcells per site11 microcells per site
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 5 The effect of microcells size for ISD = 500 m and dense user scenario
In order to determine the optimum number of microcellsper site we plot the energy efficiency and spectral efficiencygraphs with respect to the number of microcells in Figure 4Once again we observe that addition of microcells hurts theenergy efficiency for sparse user scenario For dense userscenario 8 microcells per macrocell result in the maximumenergy efficiency (see Figure 4(a)) When we observe thespectral efficiency in Figure 4(b) we conclude that 8 micro-cells per macrocell is the best choice when the microcellsize is 20 of the hexagon size It is important to note that
8 microcells per site is not an arbitrary number It is theminimum number of microcells per site that can be deployedaround the hexagon site without any gaps betweenmicrocells(see Figure 1(e))
A natural question to ask at this point is how the energyefficiency and spectral efficiency are affected by the ratio ofhexagonal site size to microcell size In a dense user scenarioand for the best energy efficiency and spectral efficiency weobserve in Figure 5 that themicrocell radius should be 30 ofthe macrocell radius when the number of microcells per site
8 Wireless Communications and Mobile Computing
8 microcells per site ISD 500
times10minus3
2
4
6
8
10
12
14
16
Ener
gy effi
cien
cy (b
itH
zJ)
20 30 40 50 60 70 80 90 10010User density
(a) Energy efficiency
8 microcells per site ISD 500
40
60
80
100
120
140
160
180
20 30 40 50 60 70 80 90 10010User density
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Figure 6 The effect of user density for ISD = 500m microcell size of 20 and the number of microcells = 8
is less than 5 On the other hand microcell radius should be20 of the macrocell radius when the number of microcellsis larger than 5 Finally we observe that the combination ofhaving 8 microcells per site and microcell size of 20 seemsto be the best scenario in terms of both energy efficiency andspectral efficiency
Finally we fix the best combination that is mentionedabove and in Figure 6 and plotted the energy efficiencyand spectral efficiency as a function of user density Weobserve that both energy efficiency and spectral efficiencyare monotone increasing functions of the user density Inthis fixed hexagonal structure idle microcells are the mainreason of having worse energy efficiency As the user densityincreases the probability of having idle microcells decreases
4 Stochastic Geometry Based Model
In this section we calculate the energy efficiency perfor-mance of stochastic geometry based deployment modelover downlink channels using Monte Carlo simulations Asdescribed in Section 2 locations of macrocell and microcellbase stations are determined according to separate Poissonpoint processes with 120582119872 and 120582120583 In order to compare theresults of the fixed grid and stochastic geometry basedmodels each 120582119872 in the stochastic geometry model is chosenso that the number of macrocells per unit area is the samein both models In a similar manner each 120582120583 is evaluatedby multiplying 120582119872 with the ratio of number of microcellsto the number of macrocells in the grid model Similar tothe fixed grid case we generate a large number of uniformlydistributed users inside the service area Then we calculateenergy efficiency and spectral efficiency of the network asa function of macrocell microcell and user densities We
assume that the coverage constraint is 95 In addition to theparameters that are given in Table 2 we use 119875120583 = minus90 dBm
In Figure 7 we demonstrate the energy efficiency per-formances of several heterogeneous scenarios with respectto macrocell density It is important to note that macrocelldensity and ISD are inversely proportional For a sparseuser case in Figure 7(a) the energy efficiency in stochasticgeometry model improves with increasing average ISD (ordecreasing macrocell density) similar to the fixed hexagonalgrid model However energy efficiency of the stochasticmodel is considerably worse than that of the grid model Forexample in a homogeneous network the energy efficiencyof the grid model is almost double of the energy efficiencyof the stochastic model In Figure 7(b) we observe that theeffect of a dense user distribution is not as drastic in stochasticmodel as it was in the grid model However deploying aconvenient number of microcells results in an improvedenergy efficiency Overall we can conclude from Figure 7that nonoverlay heterogeneous deployment has better energyefficiency in stochastic model for both sparse and dense userscenarios
Next similar to the gridmodel we investigate the effect ofour nonoverlay stochastic geometry based deploymentmodelon the area spectral efficiency Figures 8(a) and 8(b) show thatincreasing both macrocell and microcell densities improvesthe area spectral efficiency for both sparse anddense user sce-narios respectively In Figure 8(a) 8microcellssite provides2 times better area spectral efficiency than the homogeneousscenario for 120582119872 = 11547 In Figure 8(b) 11 microcellssiteprovides 3 times better area spectral efficiency than thehomogeneous scenario for 120582119872 = 11547 Thus 120582119872 is fixedat 11547 in the rest of analysis
Our next analysis is the effect of microcell density onthe energy efficiency and spectral efficiency In Figure 9(a)
Wireless Communications and Mobile Computing 9
times10minus3
16
18
2
22
24
26
28
3
32
34
36
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(a) 5 userskm2
times10minus3
2
25
3
35
4
45
5
55
6
65
7
Ener
gy effi
cien
cy (b
itsH
zJ)
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
(b) 100 userskm2
Figure 7 Energy efficiency for PPP based model
15
2
25
3
35
4
45
5
55
06 07 08 09 1 11 1205Macro density
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(a) 5 userskm2
06 07 08 09 1 11 1205Macro density
2
4
6
8
10
12
14
16
18
No micro1 micro2 micro35 micro
5 micro8 micro11 micro
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) 100 userskm2
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Figure 8 Area spectral efficiency for PPP based model
we observe that microcell addition improves the energyefficiency for both cases For sparse user case 3 micro-cellsmacrocell scenario seems to be themost energy efficientscenario whereas 11 microcellssite (120582120583 = 127017) scenariomaximizes the energy efficiency for dense user case (seeSection 2 for 120582120583 calculation) Furthermore we observe inFigure 9(b) that microcell deployment improves the spectral
efficiency for both cases Note that 8 microcellssite scenariomaximized the spectral efficiency when we employed thehexagonal grid model (see Figure 4(b)) However 11 micro-cellssite (120582120583 = 127017) scenario is themost spectral efficientscenario in Figure 9(b) It is necessary to note that thisis not a contradiction In the stochastic geometry modelmicrocell coverage regions do not necessarily overlay even
10 Wireless Communications and Mobile Computing
times10minus3
15
2
25
3
35
4
45
5
55
Ener
gy effi
cien
cy (b
itsH
zJ)
2 4 6 8 10 12 140Micro density
5 userskm2 macro density 11547100 userskm2 macro density 11547
(a) Energy efficiency
2 4 6 8 10 12 140Micro density
2
4
6
8
10
12
14
16
18
5 userskm2 macro density 11547100 userskm2 macro density 11547
Are
a spe
ctra
l effi
cien
cy (b
itss
Hz
km2)
(b) Area spectral efficiency
Figure 9 The effect of the microcell density when 120582119872 = 11547
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
Figure 10 The effect of user density when 120582119872 = 11547 and 120582120583 = 127017
if we deploy more than 8 microcells per site Thus spectralefficiency is expected to improve with the increasing numberof microcells when we use a stochastic model
Finally we fix the macrocell and microcell densities atthe most spectral efficient combination (120582119872 = 11547 and120582120583 = 127017) and consider the effect of user density As it isseen in Figure 10 increasing user density reasonably improvesboth energy efficiency and spectral efficiency It is a quiteintuitive result since the microcell utilization rate increaseswith increasing user density
5 Conclusion
In this work we investigated the effect of nonoverlay het-erogeneous network planning on the energy efficiency andspectral efficiency of cellular networks We consider twodifferent heterogeneous deployment models namely thehexagonal gridmodel and the stochastic geometrymodelWeconclude that the energy efficiency depends on the macrocelldensity microcell density and user traffic In a stochasticgeometry model that better represents the real deployment
Wireless Communications and Mobile Computing 11
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
scenarios addingmicrocells is energy efficient for even sparseuser scenarios For more structured scenarios the positiveeffects of adding microcells on the energy efficiency canonly be seen for dense user distributions Our future workis to approach this problem mathematically and propose analgorithm that finds the optimumenergy efficiency by solvingthe optimum macrocell and microcell density values for agiven user density In addition wewill considermore detailedpower consumption models in our future studies
Competing Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This work was supported by Tubitak Teydeb Grant 1505-5150007 and Ericsson The authors would like to thank MrCagatay Sagında Mr Haydar Sahin Mr Faruk Bozkurt andMr Turgut Erkul from Ericsson for their time and advice inthe preparation of this work
References
[1] Z Hasan H Boostanimehr and V K Bhargava ldquoGreen cellularnetworks a survey some research issues and challengesrdquo IEEECommunications Surveys and Tutorials vol 13 no 4 pp 524ndash540 2011
[2] L M Correia D Zeller O Blume et al ldquoChallenges andenabling technologies for energy aware mobile radio networksrdquoIEEECommunicationsMagazine vol 48 no 11 pp 66ndash72 2010
[3] C Han T Harrold S Armour et al ldquoGreen radio radiotechniques to enable energy-efficient wireless networksrdquo IEEECommunications Magazine vol 49 no 6 pp 46ndash54 2011
[4] D Feng C Jiang G Lim L J Cimini G Feng and G Y LildquoA survey of energy-efficient wireless communicationsrdquo IEEECommunications Surveys and Tutorials vol 15 no 1 pp 167ndash178 2012
[5] A Damnjanovic J Montojo Y Wei et al ldquoA survey on 3GPPheterogeneous networksrdquo IEEE Wireless Communications vol18 no 3 pp 10ndash21 2011
[6] S Tombaz K W Sung and J Zander ldquoOn metrics and modelsfor energy-efficient design of wireless access networksrdquo IEEEWireless Communications Letters vol 3 no 6 pp 649ndash6522014
[7] F Richter A J Fehske and G P Fettweis ldquoEnergy efficiencyaspects of base station deployment strategies for cellular net-worksrdquo in Proceedings of the IEEE 70th Vehicular TechnologyConference Fall (VTC rsquo09 Fall) September 2009
[8] A J Fehske F Richter and G P Fettweis ldquoEnergy efficiencyimprovements through micro sites in cellular mobile radionetworksrdquo in Proceedings of the IEEE Globecom Workshops (GcWorkshops rsquo09) Honolulu Hawaii USA November-December2009
[9] S Morosi A Fanfani and E Del Re ldquoNetwork deploymentand RRM strategies for green mobile communicationsrdquo inProceedings of the 18th European Wireless Conference (EuropeanWireless) April 2012
[10] H Ren M ZhaoW Zhou P Dong and J Kong ldquoTraffic-awaremicro base station planning in wireless cellular networksrdquo inProceedings of the IEEE 78th Vehicular Technology Conference(VTC rsquo13) IEEE Vegas Nev USA September 2013
[11] M Demirtas and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in heterogeneous cellular networksrdquo inProceedings of the IEEE 81st Vehicular Technology Conference(VTC Spring rsquo15) Glasgow UK May 2015
[12] M Demirtas and A Soysal ldquoEnergy efficient microcell deploy-ment for HetNetsrdquo in Proceedings of the 23rd Signal Processingand Communications Applications Conference (SIU rsquo15) pp 531ndash534 Malatya Turkey May 2015
[13] C Xiong G Y Li S Zhang Y Chen and S Xu ldquoEnergy- andspectral-efficiency tradeoff in downlink OFDMA networksrdquoIEEE Transactions on Wireless Communications vol 10 no 11pp 3874ndash3886 2011
[14] S Tombaz K W Sung and J Zander ldquoImpact of densificationon energy efficiency in wireless access networksrdquo in Proceedingsof the IEEE Globecom Workshops (GC Wkshps rsquo12) pp 57ndash62Anaheim Calif USA December 2012
[15] J Tang D K C So E Alsusa and K A Hamdi ldquoResourceefficiency a new paradigm on energy efficiency and spectralefficiency tradeoffrdquo IEEE Transactions onWireless Communica-tions vol 13 no 8 pp 4656ndash4669 2014
[16] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoResource allocation for energy efficiency optimization inheterogeneous networksrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 10 pp 2104ndash2117 2015
[17] J Tang D K C So E Alsusa K A Hamdi and A ShojaeifardldquoOn the energy efficiency-spectral efficiency tradeoff inMIMO-OFDMA broadcast channelsrdquo IEEE Transactions on VehicularTechnology vol 65 no 7 pp 5185ndash5199 2016
[18] M Demirta and A Soysal ldquoEnergy and spectral efficientmicrocell deployment in hetero geneous cellular networksrdquo inProceedings of the 24th Signal Processing and CommunicationApplication Conference (SIU rsquo16) May 2016
[19] J G Andrews F Baccelli and R K Ganti ldquoA tractable approachto coverage and rate in cellular networksrdquo IEEE Transactions onCommunications vol 59 no 11 pp 3122ndash3134 2011
[20] H S Dhillon R K Ganti F Baccelli and J G Andrews ldquoMod-eling and analysis of K-tier downlink heterogeneous cellularnetworksrdquo IEEE Journal on Selected Areas in Communicationsvol 30 no 3 pp 550ndash560 2012
[21] X Zhang Z Su Z Yan andWWang ldquoEnergy-efficiency studyfor two-tier heterogeneous networks (HetNet) under coverageperformance constraintsrdquo Mobile Networks and Applicationsvol 18 no 4 pp 567ndash577 2013
[22] J Peng P Hong and K Xue ldquoEnergy-aware cellular deploy-ment strategy under coverage performance constraintsrdquo IEEETransactions onWireless Communications vol 14 no 1 pp 69ndash80 2015
[23] G Auer V Giannini C Desset et al ldquoHow much energy isneeded to run a wireless networkrdquo IEEE Wireless Communi-cations vol 18 no 5 pp 40ndash49 2011
[24] 3GPP ldquoTr36814 further advancements for E-UTRA Physicallayer aspects (release-9)rdquo Tech Rep 3rd Generation Partner-ship Project 2009
[25] M S Alouini and A Goldsmith ldquoArea spectral efficiency ofcellular mobile radio systemsrdquo in Proceedings of the IEEE 47thVehicular Technology Conference May 1997
