Ameen, A. S., Berrki, D., Doufexi, A., & Nix, A. R. (2018). LTE-Advnced network inter-cell interference nlysis nd mitigtion using 3D nlogue emforming. IET Communictions, 12(13), 1563-1572. DOI: 10.1049/ietcom.2017.0765 Peer reviewed version Link to pulished version (if ville): 10.1049/iet-com.2017.0765 Link to puliction record in Explore Bristol Reserch PDF-document This is the uthor ccepted mnuscript (AAM). The finl pulished version (version of record) is ville online vi IET t http://digitl-lirry.theiet.org/content/journls/10.1049/iet-com.2017.0765. Plese refer to ny pplicle terms of use of the pulisher. University of Bristol - Explore Bristol Reserch Generl rights This document is mde ville in ccordnce with pulisher policies. Plese cite only the pulished version using the reference ove. Full terms of use re ville: http://www.ristol.c.uk/pure/out/er-terms
LTE-Advnced Network Inter Cell Interference Anlysis nd Mitigtion using 3D Anlogue Bemforming Arz Sir Ameen 1,2*, Djml Berrki 1, Angel Doufexi 1, Andrew R. Nix 1* 1 Communiction Systems & Networks Group, Deprtment of Electricl nd Electronic Engineering, University of Bristol, Bristol, United Kingdom 2 Current Affilition: Deprtment of Electricl Engineering, College of Engineering, University of Sulimni, Sulymniyh, Kurdistn, Irq * rz.meen@univsul.edu.iq ndy.nix@ristol.c.uk Astrct: This pper considers the effects of ICI on the LTE-Advnced physicl lyer downlink chnnel for different mcro cell dimeters nd se sttion (BS) ntenn heights with frequency reuse fctor of one nd three sectors per site. A site-specific 3D ry-trcing tool is used to model the communiction chnnel etween the se sttions (min nd interfering links) nd user equipment (UE) terminls. System performnce is evluted in terms of verge spectrum efficiency, cell edge throughput, nd outge proility. Two 3D nlogue emforming lgorithms re proposed to mitigte the hrmful effects of ICI. These re pplied t the BS nd/or the UE nd our results re compred with more trditionl frctionl frequency reuse deployment. Simultions demonstrte tht our proposed emforming lgorithms provide significnt system level improvements, especilly for low BS ntenn heights. With 10x10 nd 2x2 ntenn rrys t the BS nd the UE respectively, the proposed MxMin-BF lgorithms cn provide n verge spectrum efficiency of 3.6 ps/hz nd cell edge throughput of 0.56 ps/hz up to cell dimeter of 1250 m. Importntly, these results stisfy the IMT-Advnced requirements for cndidte 4G nd eyond rdio interfce technologies. Furthermore, our results outperform those chieved using frctionl frequency reuse. 1. Introduction Dt trffic demnds in cellulr networks re growing s consequence of incresing numers nd densities of User Equipment (UE) terminls nd higher dt rte pplictions. Cost nd vilility of rdio spectrum re limiting fctors for fourth genertion (4G) nd eyond cellulr Rdio Interfce Technologies (RITs). According to the requirements of the Interntionl Moile Telecommunictions-Advnced (IMT- Advnced) stndrd [1], cndidte RITs must chieve minimum downlink Averge Spectrum Efficiency (ASE) of 2.4 ps/hz/cell nd cell edge user throughput of 0.07 ps/hz/cell/user respectively. 4G RITs, such s the 3rd genertion prtnership project (3GPP) LTE-Advnced stndrd, re trgeting cellulr deployments with frequency reuse of one [2] nd three sectors per se sttion (BS) site [3]. This delivers efficient utiliztion of the licensed spectrum with the potentil for high system cpcity. A significnt increse in Inter Cell Interference (ICI) is expected in such deployments s result of Inter Site Interference (ISI) nd Inter Sector Interference (IsecI). This degrdes the Signl to Interference plus Noise power Rtio (SINR) t the UE nd hs negtive impct on ASE nd cell edge user throughput. According to [2], 10 db SINR degrdtion cn e expected with reuse fctor of one compred with reuse fctor of three. Although the incresed cpcity due to greter per-cell ndwidth lloction compenstes for the loss of cpcity due to SINR degrdtion, the cpcity of cell edge UEs is still dversely ffected. Consequently, ICI mngement nd mitigtion is vitl to ensure successful cellulr deployments. Different pproches re specified in the 3GPP LTE nd LTE-Advnced stndrds to reduce the effects of ICI in homogenous cellulr deployments. ICI Coordintion (ICIC) is introduced in 3GPP Relese 8 of LTE [4]. Relese 11 of the 3GPP stndrd includes Coordinted Multi-Point (CoMP) trnsmission nd reception [5] nd Minimum Men Squre Error Interference Rejection Comining (MMSE-IRC) [6]. These techniques require estimtion of the Chnnel Stte Informtion (CSI). The performnce evlution study in [7] showed tht the efficiency of CSI feedck re chllenges not currently considered in the stndrdistion process. Relese 12 of the 3GPP specifiction includes different network ssisted interference cncelltion techniques for the cndidte LTE-Advnced UE receivers [8], which requires communiction etween the UE nd the network. Other ville UE sed ICI mitigtion techniques include MMSE successive interference cncelltion [9] nd rte splitting inter-cell codeword cncelltion [10]. In these schemes, ICI is suppressed y sutrcting n estimte of the interference signl from the received signl depending on informtion exchnged etween the network nd UE. Other ICIC techniques include the lgorithms proposed in [11] nd [12]. In [11] the uthors proposed distriuted greedy lgorithm nd centrlized simulted nneling lgorithm to reduce the ICI in multi cell OFDMA system. These lgorithms llocte suset of the ville sucrriers to UE in given time slot for ech cell. Both lgorithms require informtion out the trffic nd the CSI from the BSs to the UEs s well s informtion exchnge mong neighoring BSs. The uthors in [12] proposed n dptive sector coloring gme (ASCG) for ICIC in homogenous network. The ASCG technique exchnges geometric network informtion (GNI) insted of chnnel stte informtion (CSI). Compred to the CSI sed ICIC, GNIsed ICIC techniques require less computtionl loding nd signling overhed ut t the cost of worst performnce. 1
ICI reduction in LTE-Advnced cellulr networks cn e chieved y using frequency reuse fctors greter thn one nd y employing Frctionl Frequency Reuse (FFR), where the sucrriers re divided into cell centre nd cell edge group [2]. The ICI is controlled y llocting different sucrriers nd trnsmit power levels to the cell centre nd the cell edge groups [13]. Compred to n equivlent deployment with frequency reuse fctor of one, lthough FFR results in incresed SINR levels (especilly for cell edge UEs), the ASE is typiclly reduced nd ccording to [14] no significnt increse is oserved in cell edge UE throughput. A power mngement technique to reduce ICI in network MIMO is proposed in [15]. The uthors showed tht the proposed power mngement comined with proper ntenn orienttion reduce the ICI nd provide improved nd consistent cpcity performnce for the UEs in the cells center nd cell edge. It is well known tht cellulr network performnce cn e incresed with the use of emforming (BF). Here liner or plnr ntenn rrys re used to chieve either orthogonlity in the chnnel mtrix or rry gin in specific direction. Digitl BF is lredy included in the 3GPP technicl specifiction for LTE-Advnced [16]. This pproch is known to e more efficient in high SINR regime since the trnsmitter must split its power cross the different sptil strems [17]. Splitting the power t low SINR results in wekening ech sptil strem which hs negtive impct on the chieved it errors nd overll system cpcity. Furthermore, cost nd complexity of digitl BF increse s the numer of ntenn elements increse since distinct RF chin is required for ech ntenn element [18]. Anlogue BF is promising lterntive technology for enhncing ASE in the low SINR regime (i.e., power nd/or interference limited scenrios), especilly t the cell edge. An rry cn e employed to increse the SINR level t the UE nd llow the use of higher order modultions. Here SINR increse ecuse of incresed signl strength nd/or reduced ICI [19]. BF weights of the ntenn rry cn e selected from pre-defined codeook sed on specific criterion. For exmple, BF in [20] is performed t the BS nd the lgorithm selects set of weights tht mximizes the received signl to noise power rtio (SNR) t the UE. Similrly, the BF in [21] lso ims to mximizes the received SNR t the UE ut the BF is employed t oth the BS nd the UE. The BF lgorithm selects BS nd UE BF weights from suset sed on hierrchicl codeooks. A mximum cpcity criterion for BS nd UE BF is proposed in [22] which mximizes the received SNR in the low SNR regime nd improves the overll cpcity for moderte nd high SNR regime. Overll cpcity is improved y utilizing prt of received SNR to enhnce the worst sucrriers in multicrrier system. In [23] n dptive interference nulling lgorithm is proposed without the need to know the direction of rrivl of the interferers. The lgorithm optimizes the weight of the ntenn rry to minimize the totl output power of the emformer while mintining the min loe gin. The ville ICI mitigtion methods in the literture require ccurte CSI t ll serving BSs. They lso require informtion exchnge etween different BSs nd different lyers of the sme BS or UE protocol hierrchy. ICI mitigtion is therefore sensitive to feedck overheds, delys nd errors. Therefore, the ojective in this pper is to nlyse the effect of ICI on the performnce of LTE- Advnced system with frequency reuse fctors of one, nd propose two nlogue BF techniques tht re insensitive to feedck overhed to mitigte the effect of ICI. The im is to stisfy the verge spectrum efficiency nd cell edge UE throughput requirements of the IMT-Advnced for cndidte 4G RITs. The ojectives of the pper cn e detiled s follows: i. To study the effect of ICI on LTE-Advnced network performnce (frequency reuse fctor of one) in relistic mcro cell scenrios tht follow 3GPP three sector hexgonl grid. A mp-sed 3D ry trcing chnnel model is used to represent the propgtion chrcteristics of the BS-UE links. The study is performed for five cell dimeters (250 m, 500 m, 750 m, 1000 m, 1250 m), three BS heights (10 m, 30 m, 50 m) t crrier frequency of 2.6 GHz in 16-squre km re in the city Centre of Bristol, United Kingdom. ii. The effect of ICI is mitigted through nlogue BF techniques tht is pplied to ll BS-UE links in the cell centre nd the cell edge using liner nd plnr rrys t the BS nd/or UE. Our BF results re compred with trditionl FFR scheme. Two BF lgorithms re proposed for use t one or oth sides of the communiction link. The methods rely on nulling the ICI signl nd incresing the min BS signl. The study is performed ssuming (10 10) plnr BS rry nd (4 4, 2 2 or 1 4) UE rry. This is to chieve the minimum downlink ASE of 2.4 ps/hz/cell nd the cell edge user throughput of 0.07 ps/hz/cell/user respectively required y IMT- Advnced for the cndidte 4G RITs. Different from the ville ICI mitigtion techniques in the literture, the proposed ICI mitigtion techniques in this pper require no informtion exchnge etween different BSs when the BF is performed t the BS side. Furthermore, there is no need for informtion feedck etween the UEs, the serving BS, nd the interfering BSs when BF is performed t the UE side. The reminder of the pper is orgnized s follows: Section 2 explins the system model including the chnnel model, network lyout nd SINR clcultion, nd nlogue emforming nd ntenn ptterns. ICI results nd nlysis re presented in Section 3 showing the impct of the BS trnsmit power nd height on the system performnce. Section 4 introduces ICI mitigtion lgorithms nd discusses the otined results. Finlly, conclusions re drwn in Section 5. 2. System Model This section presents the system model of the LTE- Advnced Physicl Downlink Shred Chnnel (PDSCH). Susection 2.1 descries the 3D ry trcing tool used to model nd generte the chnnel impulse response of ech BS- UE link. The network lyout nd the SINR clcultions re presented in susection 2.2. Susection 2.3 provides the mthemticl representtion used to generte the totl pttern of the emforming rry. The resulted pttern is then integrted with the chnnel impulse response to model the communiction chnnel etween ech BS-UE link. Finlly, susection 2.4 descries the RBIR simultion tool tht is used to evlute the throughput performnce of ech BS-UE link sed on the chnnel impulse response nd the SINR of ech BS-UE link otined from susections 2.1, 2.2, nd 2.3. 2
2.1. 3D Ry Trcing Chnnel Model The University of Bristol s outdoor ry trcer is used in this study. The model, known s Prophecy, provides pointto-point predictions of the 3D multipth for ech BS to UE link. The chnnel is modelled s set of sptil nd temporl MPCs with informtion provided on the mplitude, phse, time dely, elevtion ngle of rrivl (AoA el) nd deprture (AoD el) nd zimuth ngle of rrivl (AoA z) nd deprture (AoD z). The model identifies ll direct, trnsmitted, single scttered, doule scttered nd multiple diffrcted ry pths etween the BS nd UE. Physicl pth trcing is sed on dtse of irregulr terrin, 3D uilding nd folige structures. The model ws originlly developed in 1995 nd hs enjoyed continuously development over period of 20 yers. The model s output hs een vlidted y direct comprison with mesurements t crrier frequencies rnging from 400 MHz to 2.6 GHz [24]. Fig. 1() shows the predicted MPCs for n exmple BS-UE link. Isotropic ntenns re modelled t oth ends of the link with specific ntenn ptterns pplied s post processing sptil nd polrimetric convolution process. 2.2. Network lyout nd SINR Clcultion The LTE-Advnced system is sed on 3GPP mcro cellulr deployment with frequency reuse fctor of one. As shown in Fig. 1() ech BS site consists of three sectors with cell rdius R plced on hexgonl grid with n inter site distnce of 3R nd cell dimeter of 2R [3]. Within ech cell, nd unlike the system level simultor reported in [6], different UEs were rndomly scttered t street level. The study is performed for different cell dimeters nd BS ntenn heights t crrier frequency of 2.6 GHz in 4 km 4.4 km re in the city centre of Bristol, United Kingdom. Different BS trnsmit powers re ssumed for ech of the cell dimeters with prmeters tken from [25]. Tle 1 summrizes the system prmeters used in our studies. In Fig. 1() the min BS site lies t the centre with ICI cused y ISI nd IsecI. The ISI is generted from different sectors of the first tier of six interfering BS sites surrounding the min BS site, while the IsecI occurs from the two other sectors within the sme site s summrized in Tle 2. Given the finite set of BS loctions the resulting model generted n irregulr hexgonl grid s illustrted in Fig. 1(c). The ry trcer ws used to predict the MPCs for the min nd interfering BS-UE chnnels. The SINR t ech user loction, u, ssocited with the min BS cell, M, nd the interfering BS sites, (I: I 1, I 2,... I 6), is given y (1): Pu, M SINRu, M (1) P P AWGN P ISI u, ISI u, I seci I seci In (1), P u,m, P u,isi, P u,iseci represent the received signl strength t loction u ssocited with the min BS sector cell, ISI cells nd IsecI cells respectively. The interference power is summed cross ll interferers nd P AWGN denotes the dditive white Gussin noise, which is clculted using (2). P AWGN K T B F (2) effec liner In (2) K represents Boltzmnn s constnt, T is the temperture in Kelvin, B effec is the effective ndwidth (90% of the totl ndwidth in n LTE-Advnced OFDM system) nd F liner is the noise figure (liner vlue). In this study, 10 MHz LTE Advnced ndwidth is ssumed long with T=288 Kelvin (15 º C) nd F db=9 db [3]. Tle 1 Chnnel model prmeters Prmeter LTE Advnced Bndwidth No. of Sucrrier (N suc) No. of OFDM Symols Vlue 10 MHz 600 T slot (ms) 0.5 Crrier Frequency Environments Cellulr Deployment Minimum BS_UE distnce 7 2.6 GHz Bristol-United Kingdom 3GPP 3-sector hexgonl grid with reuse fctor of one 50 m Cell Dimeter (m) 250 500 750 1000 1250 BS trnsmit power (dbm) No. of UEs per sector 33 37 40 43 46 200 300 400 450 500 No. of BS 17 BS ntenn Height 10 m, 30 m, 50 m BS ntenn tilt 10 UE sensitivity -120 dbm Used Antenns Mcro-BS Ptch UE Antenn Bemwidth Arry elements Spcing Arry configurtion N z N el Azimuth 65º 80 360º Elevtion 15º 84 36º BS UE Liner rry 1x4 0.5 λ 2.6 GHz 10 x10 plnr Arry Tle 2 Source of ICI for different UE loctions UE loction within min BS site Sector1 Sector2 Sector3 IsecI M Sector2, M Sector3 M Sector1, M Sector3 M Sector1, M Sector2 Source of ICI Plnr Arry 2x2,4x4 ISI I 1-Secor2, I 1-Sector3, I 2-Sector3 I 3-Secor1, I 3-Sector3, I 4-Sector1 I 5-Secor1, I 5-Sector2, I 6-Sector2 3
2.3. Anlogue Bemforming nd Antenn Ptterns Liner nd plnr ntenn rrys re deployed in this study using identicl ntenn elements long with nlogue BF to enhnce the SINR in interference limited cellulr networks. In comprison with liner rry, plnr rry offers etter pttern symmetry, lower side loes nd the ility to direct the min ntenn loe to ny point in spce [26]. In the plnr ntenn rry of Fig. 2(), (N z N el) ntenn elements re plced in the (y-z) plne with n interelement spcing of d. The totl pttern of the rry is otined from (3) [26]. y N 1 N z Nz Nel el m 1 n 1 E m, n W m, n In (3), E m,n cn e clculted using (4) sed on the reference phse pttern E nd represents the pttern of ntenn element m,n plced t loction [dx m,n, dy m,n, dz m,n]. Note tht the direction of oservtion in this cse is defined y the elevtion ngle θ nd the zimuth ngle ϕ s shown in the coordinte system of Fig. 2(). Similrly, the min em cn e directed to specific direction defined y the elevtion ngle θ 0 nd the zimuth ngle ϕ 0 through djusting the weight W m,n of element m, n using (5). (3) E m, n, n 2 j ( dxm, n sin cos dym, n sin sin dzm, n cos ) E. e (4) 2 j ( dxm, n sin 0 cos 0 dym, n sin 0 sin 0 dzm, n cos 0 ) W m e (5) dx m, n 0 (6) Fig. 1. Ry trcing chnnel modelling nd network lyout c () Cptured MPCs for n exmple BS-UE link in Bristol City Centre. () The 3GPP cell topology. (c) An Exmple of UE distriution, min BS nd interfering BS loction for cell dimeter of 250 m. dy m, n ( m 1) d (7) dz m, n ( n 1) d (8) The ntenn ptterns used in this study were otined from nechoic chmer mesurements performed t the University of Bristol [27]. All ptterns we cptured in 3D nd include phse, polriztion nd directivity informtion. Since polriztion is considered in the study, the reference phse pttern E in (4) is sustituted y the elevtion nd zimuth polriztion ptterns E θ nd E ϕ respectively. The rry ptterns otined from (3) re convolved sptilly with the sptil nd temporl multipth components otined from the 3D ry trcer for ech BS-UE link. Different types nd numers of rry elements re used t ech side of the communiction link s follows: i. NoBF: BF is not pplied t either side of the link. The Mcro BS nd UE rdition ptterns shown in Fig. 2() nd Fig. 2(c) respectively re used for the links. ii. BS side BF: Fig. 2(d) shows the mesured pttern of single element in the BS rry. An rry formed from (10 10) ptch elements is deployed in the (y-z) plne. The UE pttern of Fig. 2(c) is used in this configurtion. iii. UE side BF: In this cse BF is pplied only t the UE. The BS uses the pttern shown in Fig. 2(). iv. Comined BF: BF is performed t oth ends of the link. The rry ptterns descried in cses (ii) nd (iii) ove re pplied to the BS nd UE respectively. Fig. 3() shows n exmple pttern otined from 100 ptch ntenn elements rrnged s 10x10 grid t the BS. 4
Here the ntenn weights, W m,n, {m: 0 to 9, n: 0 to 9}, re clculted using (5) with θ 0 nd ϕ 0 replced y 30 nd 45 respectively. The ntenn pttern t loction (m, n) is clculted using (4) then the totl pttern of the rry is clculted using (3). Fig. 3() shows the totl power pttern from the 2 2 rectngulr rry of UE elements. c Fig. 2. Antenn co-ordinte system nd totl power rdition ptterns () 3D Co-ordinte System. () Mcro BS ntenn totl power rdition pttern (c) UE ntenn totl power rdition pttern (d) Ptch ntenn totl power rdition pttern Fig. 3. Exmple totl power rdition ptterns of ntenn rry. () 10x10 rry of Ptch elements. () 2x2 rry of UE ntenn elements. 2.4. RBIR Link Level Astrction The performnce of the LTE-Advnced PDSCH is evluted for five cell dimeters nd three BS ntenn heights. Performing it ccurte physicl lyer simultions to estimte sttisticlly relevnt system level performnce for lrge numers of BS-UE links nd for mny different Modultion nd Coding Schemes (MCS) is time consuming. Insted, the Received Bit mutul Informtion Rte (RBIR) strction technique [28] cn e used s computtionlly d efficient lterntive to it level simultion. In [29] we reported excellent greement etween it level simultion nd RBIR strction, with the ltter running round 300 times fster on the sme computing pltform. The comined chnnel nd ntenn impulse response for ech serving BS-UE link is converted into the frequency domin nd used s the input into our PDSCH RBIR strction engine to estimte the instntneous pcket error rte (PER) for 10 MCS modes t the verge SINR determined y (1). A link dpttion lgorithm is pplied to select the MCS mode tht mximizes the dt throughput (THR) of ech link. The LTE-Advnced PDSCH throughput is clculted using (9) [30] ssuming perfect chnnel knowledge t the receiver. No reference signls re used to estimte the communiction chnnel t the receiver. THR MCS R 1 PER (9) MCS MCS In (9) R MCS represents the pek error free dt rte for the considered MCS mode which cn e clculted using (10), nd PER MCS is the chieved PER for the considered MCS determined using the PDSCH RBIR strction engine. R MCS is function of modultion order (k m), the coding rte (R c), the numer of ctive sucrriers (N suc) nd the numer of OFDM symols (N sym) in the time slot (T slot). Tle 1 summrizes the system prmeters used here while Tle 3 lists the vlue of R MCS for ech considered MCS mode. km Rc Nsuc Nsym RMCS (10) Tslot Tle 3 List of MCS Modes nd Pek Error Free Dt Rtes MCS Modultion Code rte R MCS (Mps) for SISO 1 1/3 5.6 2 QPSK 1/2 8.4 3 [k m=2] 2/3 11.2 4 4/5 13..44 5 1/2 16.8 6 16QAM [k m=4] 2/3 22.4 7 4/5 26.88 8 2/3 33.6 9 64QAM [k m=6] 3/4 37.8 10 4/5 40.32 3. ICI Results nd Anlysis This section presents the LTE-Advnced PDSCH simultion results for single strem for theoretic deployments in Bristol t crrier frequency of 2.6 GHz. The results re presented using metrics such s SNR, SINR, pth loss, Line of Sight (LoS) proility, ASE, cell edge throughput, nd UE outge proility. Susection 3.1 shows the impct of BS trnsmit power on the ASE. The impct of the BS ntenn height on the SINR nd SNR is nlysed in susection 3.2, while susection 3.3 presents the impct of the BS ntenn height on the ASE, cell edge UE throughput, nd the user outge proility. These prmeters re defined in [1] s follows: Averge Spectrum Efficiency is mesured in it/second/hz/cell (ps/hz/cell) nd defined s the ggregte throughput for ll users normlised y the overll cell ndwidth nd the numer of cells. 5
Cell edge UE throughput represents the 5% point on the cumultive distriution function (CDF) of the UE throughput normlised y the totl cell ndwidth. User outge proility: A UE is considered to e in outge if its throughput drops to zero. Note tht the pth loss is clculted per UE from the ry trcing predictions nd the men pth loss models shown in this section re included to support our conclusions. 3.1. Impct of BS Trnsmit Power on ASE This section studies the effect of the BS trnsmit power on ASE to justify the selected vlues of the BS trnsmit power for different cell dimeters in Tle 1. The study is performed for BS trnsmit power rnge from 20 dbm to 50 dbm considering ICI. The ASE versus BS totl trnsmit power for different BS ntenn heights nd cell dimeter of 750 m is shown in Fig. 4() while Fig. 4() considers different cell rdiuses for BS ntenn height of 30 m. Both figures show tht the selected BS totl trnsmit power (Pt) in the study provide t lest 90% of the chievle ASE when the totl trnsmit power set to 50 dbm. This confirms the proper selection of the BS trnsmit power for different cell dimeters sed on [25]. Any increse in the totl trnsmit power eyond the selected vlues in Tle I hve no significnt impct on the system performnce. 3.2. Impct of BS Antenn Height on SINR nd SNR The effect of BS ntenn height on the SNR nd SINR sttistics is shown in Fig. 5() for cell dimeter of 750 m. As expected, the SNR level increses with incresing BS ntenn height. This is the result of lower pth loss to the UEs s the BS height increses (see Fig. 5()). A mjor contriutor to the lower pth loss is the incresed LoS proility for the min BS-UE links (rising from 3.6% nd 5.4% for BS heights of 10 m nd 30 m respectively to 12.2% for height of 50 m (see Tle 4). In contrst to the ove, the SINR grphs presented in Fig. 5() show very different trend. The inter site interfering BSs with the highest ntenn heights experience the lowest pth loss vlues (shown s dshed lines in Fig. 5()). This hs negtive impct on the SINR level. As reported in Tle 4, incresed ISI with incresed interfering BS ntenn height is lso result of incresed LoS proilities. For BS ntenn heights of 30 m nd 50 m the pth loss to the ISI BSs is slightly more thn the corresponding min BS. This is not the cse with BS ntenn height of 10 m, where the interfering BS-UE pth loss is significntly higher thn the min BS-UE pth loss. Fig. 4. ASE versus BS Trnsmit Power for different BS ntenn heights nd cell dimeters (CD). () Different BS ntenn heights, cell dimeter =750 m () Different cell dimeters with BS ntenn height of 30 m Fig. 5. SNR, SINR nd men pth loss model for 3GPP deployment (different BS ntenn heights, cell dimeter of 750 m) () SINR nd SNR () Best Fit Pthloss Model 6
Without ICI With ICI For min BS-UE link, ech of its interfering links is considered to e in outge if the power sum of ll interfering MPCs is less thn the UE sensitivity (ssumed to e 120 dbm in this study). As reported in Tle 4, the outge proility for 10 m BSs is 65.2%. This drops to 27.6% nd 13.4% for interfering BS heights of 30 m nd 50 m respectively. Fig. 5() lso shows slight differences in the pth loss of IsecI BS-UE links for different BS heights. Overll, s confirmed y Fig. 5(), the SINR sttistics re etter for 10 m BS heights nd degrde with incresing BS height. Tle 4 List of LoS nd ISI Link Outge Proilities for Cell Dimeter of 750 metres Prmeter LoS Proility of Min BS-UE links LoS Proility of ISI BS-UE links ISI BS-UE links Outge Proility BS Antenn Height (m) 10 30 50 3.6 % 5.4 % 12.2 % 0.9 % 1.4 % 3.4 % 65.2 % 27.6% 13.4% 3.3. Impct of BS Antenn Height on ASE, Cell Edge Throughput nd Outge Proility The ASE is compred for different cell dimeters nd BS ntenn heights in Fig. 6(). It is cler from the figure tht when ICI is considered 10 m BS height offers the est ASE for ll cell dimeters with mximum ASE vlue of 2 ps/hz/cell. In contrst, the ASE grphs without the effect of ICI clerly shows etter performnce for higher BS ntenn heights. Considering the 10 m BS ntenn height, Fig. 6() lso shows tht ASE decreses s the cell dimeter increses. In ddition, Fig. 6() shows the CDFs of the UE normlized throughput for different cell dimeters. The 5% UE throughput corresponds to zero for ll cses, which mens tht the cell edge throughput is equl to zero. The chieved vlues of the ASE nd cell edge throughput considering ICI re less thn the IMT-Advnced requirement of 2.4 ps/hz/cell nd 0.07 ps/hz/cell/user respectively. Therefore, SINR enhncement is vitl to improve the ASE nd cell edge throughput. For this purpose, two different nlogue BF techniques re proposed in Section 4. These re then compred ginst the FFR technique. Tle 5 lists the cell edge throughputs nd outge proilities for different cell dimeters nd BS ntenn heights for two different cses; with nd without the effects of ICI. In oth cses the outge proility increses with incresing cell dimeter with the lowest vlue oserved for cell dimeter of 250 m. Unlike the ICI cse, higher cell edge throughputs nd lower outge proilities re oserved for 50 m BS ntenn heights compred to the 10 m ntenn height when neglecting ICI. Fig. 6. Averge spectrum efficiency nd CDF of UE throughput () Averge spectrum efficiency for different BS ntenn heights nd cell dimeters with nd without ICI. () CDF of UE throughput for different cell dimeters (CD) for 10 m BS ntenn height with ICI. Tle 5 List of Cell Edge Throughput nd Outge Proility Cell Dimeter (m) Cell Edge UE Throughput (Mps/Hz) UEs Outge Proility H10 H30 H50 H10 H30 H50 250 0 0 0 0.11 0.20 0.20 500 0 0 0 0.14 0.30 0.30 750 0 0 0 0.15 0.35 0.33 1000 0 0 0 0.22 0.44 0.40 1250 0 0 0 0.27 0.53 0.44 250 0.68 3.41 3.96 0.02 0.00 0.00 500 0.00 0.00 1.38 0.06 0.06 0.01 750 0.00 0.00 0.17 0.08 0.08 0.04 1000 0.00 0.00 0.00 0.13 0.13 0.07 1250 0.00 0.00 0.00 0.17 0.17 0.08 7
4. ICI Mitigtion Next, we investigte the use of nlogue emforming to reduce the effects of ICI in n LTE-Advnced cellulr network. The ntenn rry is defined y N z N el elements rrnged uniformly in the elevtion plne. BF is performed t the BS using 10 10 rry nd t the UE side using 1 4, 2 2, or 4 4 rrys. The study is performed ssuming the interfering BSs do not exchnge informtion with the serving BS. Furthermore, when considering BF t the min BS we ssume rndom BF t the interfering BS sites. As explined in Section 2.3, BF is performed in the min cell y pplying weights to the ntenn elements. These weights re clculted using (5) s function of the ntenn element position nd the desired steering direction descried y θ 0 nd ϕ 0 in the elevtion nd zimuth plnes respectively. Three different lgorithms re used to clculte the rry weights for SINR enhncement. The lgorithms either 1) increse the received signl strength, 2) null the ICI or 3) perform mixture of the two s descried in susections 4.1, 4.2, nd 4.3 respectively. The performnce of the three BF lgorithms is lso compred with the FFR technique discussed in susection 4.4. Finlly, susection 4.5 shows the impct of the rry configurtion on the system performnce nd the reltion etween the throughput nd other chnnel prmeters. 4.1. Mximum Ry Power BF (MxRP-BF) In this lgorithm, the vlues of θ 0 nd ϕ 0 in (5) re sustituted y the deprture or rrivl ngles of the mximum power MPC t the BS or UE respectively s summrized in Algorithm 1. As mentioned in Section 2.1, the ngulr informtion of the mximum power MPC of ech BS-UE link is otined from the ry trcer tool, however prcticlly this cn e estimted using specil techniques. The effectiveness of this pproch in terms of incresing the serving BS signl strength is investigted t oth the BS nd UE. Fig. 7() shows the ASE for the different BF lgorithms nd for numer of cell dimeters. Results without BF nd ICI re lso included in the figure for comprison. As mentioned in Section 3, etter performnce is oserved for BS ntenn heights of 10 m when ICI is included in the simultion. In the ICI free cse 50 m BS height yielded the est performnce. Hence, the results in this section ssume BS height of 10 m in ll cses other thn the interference free scenrio, where 50 m BS height is pplied. Algorithm 1: Procedure of MxRP-BF 1 Find the mximum power MPC 2 if BS side BF: θ 0 = elevtion deprture ngle of the mximum power MPC ϕ 0 = zimuth deprture ngle of the mximum power MPC else if UE side BF: θ 0 = elevtion rrivl ngle of the mximum power MPC ϕ 0 = zimuth rrivl ngles of the mximum power MPC 3 Clculte weight W m,n using (10) for ech ntenn element Compring the ASE grphs of Fig. 7(), MxRP-BF t the BS exceeds the IMT-Advnced requirement of 2.4 ps/hz. Furthermore, the performnce of MxRP-BF t the UE is close to the NoBF cse nd lower thn the IMT- Advnced requirement for ll cell dimeters. MxRP-BF is more effective t the BS for two resons. Firstly, considering the power CDF grphs in Fig. 7(), MxRP-BF t the BS results in higher totl received power thn MxRP-BF t the UE. This is result of the higher directivity gin otined from the 10 10 rry t the BS (see Fig. 3()) compred to the 4 4 rry t the UE. Secondly, MxRP-BF t the UE leds to ICI enhncement in ddition to the min signl power, while the ppliction of rndom BF t the interfering BSs reduces the ICI power t the UE. 4.2. Proposed Minimum ICI BF (MinICI-BF) This section proposes novel lgorithm for selecting the BF weights for nlogue BF. The ville techniques in the literture enhnce the system performnce y incresing the received signl strength t the UE [20]- [22]. In contrst our proposed MinICI-BF lgorithm ims to minimize the totl received ICI power y pplying predefined set of weights tht provide the highest SINR vlue (see lgorithm 2). This lgorithm is proposed for exclusive use t the UE, where informtion exchnge with ny BS (including the min BS) is not required. Different weights re clculted y pplying different vlues of θ 0 nd ϕ 0 in increments of Δx, where Δx is ssumed to e 30 for plnr rry nd 10 for liner rry. The ASE results for MinICI-BF nd for different rry configurtions re presented in Fig. 7(). The results show performnce increse compred to the cses of UE MxRP- BF nd NoBF. For MinICI, the performnce of 4 4 is greter thn the 2 2 nd 1 4 rrys. The 2 2 nd 1 4 rrys fil to chieve the ASE requirements of 2.4 ps/hz for cell dimeters of 500 m nd ove. The 4x4 rry cn provide n ASE of 2.4 ps/hz for cell dimeters up to 750 m. Focusing on the performnce difference etween plnr nd liner rry elements (keeping the totl numer of elements constnt), our CDF plots of the received min nd ICI power in Fig. 7(c). show tht the plnr rry performs slightly etter thn the liner rry. The CDF dt shows slightly lower level of received ICI power when the plnr rry is used. It cn e seen from Fig. 7(c) tht the ility of the MinICI-BF lgorithm to reduce ICI is etter thn its ility to increse the min signl strength. This explins why the 1 4 nd 2 2 MinICI-BF outperforms the 4 4 MxRP-BF t the UEs despite the higher rry gin with more ntenn elements. This explins the ility of our proposed MinICI-BF lgorithm to increse system performnce. However, the performnce is worse thn the interference-free scenrio. To see the impct of the different BF lgorithms on the cell edge users, Tle 6 list the cell edge throughput for ech lgorithm for the sme scenrios s those depicted in Fig. 7(). As mentioned in the introduction, the IMT-Advnced trget for cell edge throughput is 0.07 ps/hz. It is cler from Tle 6 tht mong the forementioned lgorithms the cell edge throughput trget is only stisfied with i) the 10 10 BS rry with the MxRP-BF lgorithm for cell dimeters up to 1000 m nd ii) the 4 4 UE rry with the MinICI-BF lgorithm for cell dimeter of 250 m. In the following section our proposed MinICI-BF pproch, which is pplied 8
exclusively t the UE side, is extended y lso dding MxRP-BF t the BS. 4.3. Comined MxRP nd MinICI BF Given the effectiveness of the MxRP-BF nd MinICI-BF lgorithms to enhnce SINR when pplied t the BS nd UE respectively, further system performnce cn e chieved y running oth lgorithms concurrently in new lgorithm (MxMin-BF). The BS performs MxRP-BF to increse received min received power t the UE while the UE minimizes the ICI power through MinICI-BF s summrized in Algorithm 3. In Fig. 7() the MxMin-BF results re presented for 10 10 BS ntenn rry nd 2 2 nd 1 4 UE rrys. The ASE in oth cses outperforms the interference-free results for cell dimeter of 750 m nd ove. Very close results re lso oserved for cell dimeter of 500 m. Fig. 7(d) shows tht the SINR level of the MxMin-BF lgorithm exceeds the interference-free scenrio. Moreover, from Tle 6 it cn e seen tht the IMT-Advnced cell edge throughput requirement is stisfied for ll cell dimeters when 2 2 nd 1x4 UE rrys re deployed. Algorithm 2: Procedure of MinICI-BF 1 Determine codeook of ntenn weights: For θ 0 =0 to 180 in steps of Δx For ϕ 0 =0 to 360 in steps of Δx Clculte BF weights W m,n using (10) for ech ntenn element (m,n) 2 For ech set of BF weights in the codeook: Clculte received SINR t the UE Choose the BF weight tht results in the lowest received interfering power nd mximize the received SINR. Algorithm 3: Procedure of MxMin-BF 1 Apply MxRP-BF (Algorithm 1) t the BS side. 2 Apply MinICI-BF (Algorithm 2) t the UE side. c Tle 6 List of Cell Edge Throughputs for Different BF Algorithms Cell Edge throughput (ps/hz) Cse Cell Dimeter (m) 250 500 750 1000 1250 No-BF, With ICI 0 0 0 0 0 MxRP-BS10 10 0.748 0.418 0.392 0.090 0 MxRP-UE4 4 0 0 0 0 0 MinICI-UE4 4 0.297 0 0.004 0 0 MinICI-UE2 2 0.000 0 0 0 0 MinICI-UE1 4 0 0 0 0 0 MxMin, BS10 10, UE2 2 1.982 1.090 1.078 0.643 0.562 MxMin, BS10 10, UE1 4 1.698 0.913 0.888 0.492 0.336 No-BF, No ICI 1.210 0.512 0.534 0.002 0 FFR 0.04 0 0 0 0 d Fig. 7. Performnce of different BF lgorithms BS ntenn height of 10 m. () Averge Spectrum Efficiency for different cell dimeters () CDF of the received min nd ICI powers of MxRP-BF nd No-BF for cell dimeter of 500 m. (c) CDF of the received min nd ICI powers of MinICI-BF nd No-BF for cell dimeter of 500 m. (d) CDF of the SINR for different simultion scenrios for cell dimeter of 750 m. 9
4.4. Comprison with Three Sector FFR Approch In this section the performnce of the proposed BF lgorithms is compred ginst the three-sector FFR deployment descried in [31]. Ech sector is divided into cell centre nd cell edge groups. In this study 50% of the ville rdio resources re ssigned to the cell centre group with the remining divided eqully etween the three cell edge group sectors. Furthermore, the centre rnge is ssumed to extend to 60% of the cell edge rnge. The selection of these FFR prmeters is sed on recommendtions in [31] for optiml performnce. It is cler from Fig. 7() tht the ASE for the FFR scenrio is lower when compred with the other scenrios, including the NoBF cse, due to the use of spectrum prtitioning. The outge proility results in Fig. 8 show significntly etter performnce for the FFR technique compred with MxRP-BF performed t the UE nd the NoBF cse for ll cell dimeters. FFR outge performnce is lso slightly etter thn the MinICI with 1 4 rrys performed t the UE for cell dimeter of 250 m. However, ll other BF lgorithms outperform the FFR scenrio. The cell edge throughput results in Tle 6 show tht the use of FFR is unle to stisfy the cell edge throughput requirement of IMT-Advnced for ll cell dimeters. This confirms the conclusion of [14]. Finlly, we compre the SINR grphs of Fig. 7(d) to the FFR scenrio with the MinICI-BF lgorithm using 2 2 UE rry. By comining 2 2 UE rry with the MinICI-BF lgorithm it is possile to exceed the SINR levels using FFR. SINR vlue. Fig. 9() shows the CDF grphs of SINR level t the UEs for the different rry configurtions of the MinICI lgorithm. It is cler from the figure tht the MinICI lgorithm in enhncing the SINR s the numer of ntenn elements in the rry increses. The SINR enhncement is due to reduction in the received ICI power comined with the increse in the totl received power from the serving BS. The BF lgorithm increses the received power from the serving BS in specific direction while reducing the received power from other directions. This results in received multipth component (MPCs) with high power dominnt MPC compred to the other MPCs. This in turn leds to increse in the K fctor nd decrese in the RMS DS with the increse in the numer of elements in the rry s shown in the CDF grphs of the K fctor nd the RMS DS in Fig. 9() nd Fig. 9(c) respectively. Fig. 9(d) show tht the throughput t the UE increses s numer of the elements in the rry increses. This increse in the throughput is contriuted to the increse of the SINR levels, increse in the K fctor, nd the decrese in the RMS DS t the UE s the numer of elements in the rry increses from one element in the No-BF cse, 4 elements in the 2x2 nd 1x4 cses, to 16 elements in the 4x4 cse. Finlly, to compre etween the impct of the liner nd plnr rry rrngement on the performnce, we consider the 2x2 nd the 1x4 UE rry configurtions where in oth cses the numer of ntenn elements in the rry is 4. It is cler from the grphs of Fig. 9 tht the plnr rry provides slightly etter performnce thn the liner rry. This is ecuse the liner rry pplies BF in one plne only (elevtion or zimuth) while the plnr rry performs BFs in oth plnes. Fig. 8. Outge proility for different BF lgorithms nd cell dimeters for BS ntenn height of 10 m. 4.5. Impct of the Arry Configurtion on the Performnce This section investigtes the impct of the numer of ntenn elements in the BF rry nd their rrngement s liner or plnr rry on the UE throughput, the SINR level t the UE, the K fctor, nd the Root Men Squre (RMS) Dely Spred (DS). We lso determine reltion etween the throughput t the UE nd the other forementioned chnnel prmeters. For this nlysis we consider the different rry configurtions of the proposed MinICI lgorithm tht is pplied t the UE using 4x4, 2x2, nd 1x4 rry configurtions. The throughput nd the chnnel prmeters for the No-BF cse re lso included in the nlysis for comprison nd reference. As mentioned previously, the proposed MinICI-BF lgorithm ims to minimize the totl received ICI power y pplying predefined set of weights tht provide the highest 10
c d Fig. 9. Performnce of MinICI BF lgorithm for different rry configurtion with BS ntenn height of 10 m. () CDF of the SINR level t the UEs. () CDF of the K fctor t the UEs. (c) CDF of RMS Dely Spred (DS) t the UEs. (d) CDF of the of UEs Throughput. 5. Conclusions In this pper, the performnce of n LTE-Advnced homogenous network deployment ws evluted ssuming frequency reuse fctor of one in terms of verge spectrum efficiency, cell edge throughput nd outge proility using PDSCH strction engine simultor. The communiction chnnel etween the min BS-UE links nd the interfering BS-UE links were modelled using site-specific mp-sed 3D ry-trcing tool sed on relistic city-centre scenrio in Bristol. The study ws performed for vrious cell dimeters, nd BS ntenn heights t crrier frequency of 2.6 GHz. Liner nd plnr rrys were deployed t the BS nd UE sides of the link with two proposed BF lgorithms pplied to increse system performnce. Results were lso compred ginst previously reported three-sector FFR technique. The following conclusions cn e drwn: When ISI nd IsecI re considered, etter network performnce ws chieved for BS heights of 10 m (compred with 30 m nd 50 m). This occurred ecuse of the higher pth loss experienced y the interfering BS- UE links compred to the min BS-UE links t lower BS heights. Plnr ntenn rrys nd nlogue emforming t the BS nd UE represent promising technology to meet the cpcity requirements of future networks. The enefits re prticulrly compelling for cell edge users, where there is no requirement to increse BS trnsmit powers or to prtition the ville spectrum. The increse in the numer of the ntenn elements in the rry provides improved effective K fctor nd RMS dely spred s well s incresed throughput nd SINR level t the UE. Unlike other ICI mitigtion methods (including BS controlled BF), the MinICI-BF lgorithm does not require informtion feedck etween the UEs nd the BSs. Given the dvncements eing mde in digitl signl processing nd electronic circuit friction it is now possile to integrte 2 2 rrys into smrtphone UEs nd 4 4 rrys into tlet UEs. 6. Acknowledgments Arz Sir Ameen would like to thnk the University of Sulimni nd the HCDP directorte t the Ministry of Higher Eduction nd Scientific Reserch in Kurdistn of Irq for sponsoring his PhD study. 7. References [1] 3GPP TR36.913-V10: 'Requirements for Further Advncements for Evolved Universl Terrestril Rdio Access; (LTE Advnced-Relese10) ', 2011. [2] Himyt, N., Tlwr, S., Ro, A., et l.: 'Interference Mngement for 4G Cellulr Stndrds', IEEE Communictions Mgzine, 2010, 48, (8), pp. 86-92. [3] 3GPP TS36.942-V10.2: 'Evolved Universl Terrestril Rdio Access: Rdio Frequency System Scenrios', 2010. [4] 3GPP TS36.300-V10.7: 'Evolved Universl Terrestril Rdio Access nd Evolved Universl Terrestril Rdio Access Network: Overll Description; Stge2', 2012. [5] 3GPP TR36.819-V11.2: 'Coordinted Multi-Point Opertion for LTE Physicl Lyer Aspects', 2013. [6] 3GPP TR36.829-V11.1: 'Enhnced Performnce Requirement for LTE User Equipment', 2012. [7] Sun, S., Go, Q., Peng, Y., et l.: 'Interference Mngement Through CoMP in 3GPP LTE-Advnced Networks', IEEE Wireless Communictions Mgzine, 2013, 20, (1), pp. 59-66. [8] 3GPP TR36.866-V12: 'Study on Network Assisted Interference Cncelltion for LTE', 2014. [9] Hrdouin, E., Hssn M., Sdni A.: 'Downlink Interference Cncelltion in LTE: Potentil nd Chllenges', Proc. Int. Conf. Wireless Communictions nd Networking Conference(WCNC), Shnghi, Chin, April 2013, pp. 3597-3602. [10] Zhou, G., Xu, W. Buch, G.: 'Network Assisted Inter- Cell Codeword Cncelltion for Interference-Limited LTE-A nd Beyond', Proc. Int. Conf. Wireless Communictions nd Networking Conference Workshops (WCNCW), Istnul, Turkey, April 2014, pp. 52-57. [11] Gupt, V., Nmir, A., Ksekr, G.: 'Complexity Anlysis, Potentil Gme Chrcteriztion nd 11
Algorithms for the Inter Cell Interference Coordintion with Fixed Trnsmit Power Prolem', IEEE Trnsctions on Vehiculr Technology,2017, DOI: 10.1109/TVT.2017.2774862. [12] Lee, W., Choi, H., Kim, H., et l.: 'Adptive Sector Coloring Gme for Geometric Network Informtion- Bsed Inter-Cell Interference Coordintion in Wireless Cellulr Networks', IEEE/ACM Trnsctions on Networking, 2018, 26, (1), pp. 288-301. [13] Qin, M., Hrdjwn, W., Li, Y., et l.: 'Adptive Soft Frequency Reuse Scheme for Wireless Cellulr Networks', IEEE Trnsctions on Vehiculr Technology, 2015, 64, (1), pp. 118-131. [14] Boudreu, G., Pnicker, J., Guo, N., et l.: 'Interference Coordintion nd Cncelltion for 4G Networks', IEEE Communictions Mgzine, 2009, 47, (4), pp. 74-81. [15] Sheu, J., Lyu, S., Hung, C.: 'On ntenn orienttion for inter-cell interference coordintion in cellulr network MIMO systems, Journl of Communictions nd Networks, 2016, 18, (4), pp. 639-648. [16] 3GPP TS36.211-V10.4: 'Evolved Universl Terrestril Rdio Access: Physicl Chnnels nd Modultion', 2011. [17] Sun, S., Rppport, T., Heth, R., et l.: 'MIMO for Millimeter-Wve Wireless Communictions: Bemforming, Sptil Multiplexing, or Both?', IEEE Communictions Mgzine, 2014, 52, (12), pp.110-121. [18] Sohri, F., Yu, W.: 'Hyrid Digitl nd Anlog Bemforming Design for Lrge-Scle Antenn Arrys', IEEE Journl of Selected Topics in Signl Processing, 2016, 10, (3), pp. 501-513. [19] Bi, T., Alkhtee, A., Heth, R.: 'Coverge nd Cpcity of Millimeter-Wve Cellulr Networks', IEEE Communictions Mgzine, 2014,52, (9), pp. 70-77. [20] Choi, J.: 'Anlog Bemforming for Low-Complexity Multiuser Detection in mm-wve Systems', IEEE Trnsctions on Vehiculr Technology, 2016, 65, (8), pp. 6747-6752. [21] Hur, S., Kim, T., Love, D., et l.: 'Millimeter Wve Bemforming for Wireless Bckhul nd Access in Smll Cell Networks', IEEE Trnsctions on Communictions, 2013, 61, (10), pp. 4391-4403. [22] Vi, J., Sntmri, I., Elvir, V., et l.: 'A Generl Criterion for Anlog Tx-Rx Bemforming Under OFDM Trnsmissions', IEEE Trnsctions on Signl Processing, 2010, 58, (4), pp. 2155-2167. [23] M, X., Lu, L., Sheng, W., et l: 'Adptive interference nulling with pttern mintining under min loe suspce nd qudrtic constrints, IET Microwves, Antenns & Propgtion, 2018, 12, (1), pp. 40-48. [24] Tmeh, E., Nix, A.: 'The Use of Mesurement Dt to Anlyse the Performnce of Rooftop Diffrction nd Folige Loss lgorithms in 3-D Integrted Urn/Rurl Propgtion Model', Proc. Int. Conf. Vehiculr Technology Conference (VTC), Ottw, Cnd, My 1998, pp. 303-307. [25] 'High-Cpcity Indoor Wireless Solutions: Picocell or Femtocell?', www.fujitsu.com/us/imges/high-cpcity-indoor- Wireless.pdf, ccessed April 2015. [26] Blnis, C.: 'Arrys: Liner, Plnr, nd Circulr', in: 'Antenn Theory: Anlysis nd Design' (John Wiley & Sons, New York, 1997, 2nd edn.), pp. 249-320. [27] Mellios, E., Mnsor, Z., Hilton, G., et l.: 'Impct of Antenn Pttern nd Hndset Rottion on Mcro-cell nd Pico-cell Propgtion in Heterogeneous LTE networks', Proc. Int. Symp. Antenns nd Propgtion, Chicgo, USA, July 2012. [28] Wn, L., Tsi, S., Almgren, M.: 'A Fding-Insensitive Performnce Metric for Unified Link Qulity Model', Proc. Int. Conf. Wireless Communictions nd Networking Conference (WCNC), Ls Vegs, USA, April 2006, pp. 2110-2114. [29] Ameen, A., Mellios, E., Doufexi, A., et l.: 'LTE Advnced Downlink Throughput Evlution in the 3G nd TV White Spce Bnds', Proc. Int. Symp. Personl Indoor nd Moile Rdio Communictions (PIRMC), London, UK, Septemer 2013, pp. 771-775. [30] Beh, K., Doufexi, A., Armour, S.: 'Performnce Evlution of Hyrid ARQ Schemes of 3GPP LTE OFDMA System', Proc. Int. Symp. Personl Indoor nd Moile Rdio Communictions (PIMRC), Athens, Greece, Septemer 2007, pp. 1-5. [31] Squi, N., Hossin, E., Kim, D.: 'Frctionl Frequency Reuse for Interference Mngement in LTE-dvnced HetNets', IEEE Wireless Communictions Mgzine, 2013, 20, (2), pp. 113-122. 12