ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons Assessment of LTE Uplnk Power Control wth Dfferent Frequency Reuses Schemes Mohamed M. El-Ghawaby Electroncs and Communcatons AAST Caro, Egypt ghawaby25@gmal.com Hesham El-Badawy Network Plannng Department Natonal Telecom Insttute Caro, Egypt heshamelbadawy@ieee.com Hazem H. Al Electroncs and Communcatons AAST Caro, Egypt hazemhal@gmal.com Abstract Sngle Carrer Frequency Dvson Multple Access (SC-FDMA) s the access scheme chosen by 3GPP for uplnk UTRAN Long Term Evoluton project (LTE). As SC-FDMA provdes ntra-cell orthogonalty, one of the man reasons for performance degradaton s the Inter-Cell Interference (ICI). Ths degradaton s accentuated by the frequency reuse of deployed n the system, Snce the Frequency Reuse (FR) and Power Control (PC) functonaltes s a strong tool for cochannel nterference mtgaton, usng them crtcal ssues n cellular Orthogonal Frequency Dvson Multple Access (OFDMA)/LTE networks. In ths paper, we compare between the Open Loop Power Control (OLPC) and Closed Loop Power Control (CLPC) performance usng dfferent frequency reuse schemes. Smulaton results show that large dfferences exst between the performance of dfferent (FR) schemes and the optmal case n the overall cell throughput, as well as the cell-edge user performance. Also the closed loop power control has shown more cell and edge throughput gan over OLPC. Keywords Open Loop Power Control; Closed Loop Power Control; Hard Frequency Reuse; Fractonal Frequency Reuse; Soft Frequency Reuse. I. INTRODUCTION LTE ntroduces a number of nnovatons that, n aggregate, contnue to push ever closer to the theoretcal maxmum data rates defned by Shannon's Law [4]. Advances n mult-antenna technques, OFDMA methods, wder bandwdth, nterference mtgaton, and protocol effcences are fundamental to delver the promse of 4G Mass Market Wreless Broadband. The amazngly hgh data rates and sector throughputs (capacty) per cell are fundamental to supplyng the ever ncreasng demand for wreless broadband. Effectve reuse of resources n a cellular system can hghly enhance the system capacty. Wth a smaller Frequency Reuse Factor (FRF), more avalable bandwdth can be obtaned by each cell. So, n ths sense the classcal FRF of s desrable see Fg. 2a. However, wth the usage of FRF-, the most User Equpments (UEs) are serously afflcted wth heavy ICI, especally near the cell edge. And that causes severe connect outages and consequently low system capacty. The conventonal method to fgure out ths problem s through ncreasng the cluster-order, whch can mtgate the ICI effcently, nonetheless at the cost of a decrease on avalable bandwdth for each cell. Ths leads to restrcted data transmssons and lower system spectrum effcency. To take am at mprovng cell-edge performance whle retanng system spectrum effcency of reuse-. There are many technques whch can be used to mtgate nterference n E-UTRA uplnk. The basc approaches are classfed nto dfferent type such as Power Control, Inter-cell-nterference randomzaton, Coordnaton/avodance, and Frequency doman spreadng. Recent researches are focused at OLPC and CLPC performance evaluaton. Ths s due to ts capablty of nterference mtgaton as well as ncreasng the system throughput. Many nvestgatons for the performance and confguratons of the OLPC and CLPC [4][7]. Results show that the dfferent confguraton s drectly effect on both cell edge users and cell center users. Also, many recent researches are focused at FR technques such as Hard Frequency Reuse HFR, Fractonal Frequency Reuse FFR, Soft Frequency Reuse SFR and performance evaluaton and developng [5][6], Results show great performance, especally for the cell edge throughput due to nterference mtgaton. Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 2
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons The current paper nvestgates about the ICI as a result of uplnk PC and FR. In addton, t wll combne between each PC technques and the three FR schemes to acheve better performance. The paper s organzed as follows; Secton II descrbes the general nterference mtgaton concepts for E-UTRA followed by detal descrpton of OLPC, CLPC and the most famous frequency reuse schemes whch wll be used wth both OLPC & CLPC. Secton III s dscussng proposed system model. Secton IV llustrates results and ts analyss. Fnally, the concluson s presented n Secton V. II. INTERFERENCE MITIGATION PC and FR schemes are representng the man buldng blocks of the proposed system model. A. Open Loop Power Control: PC refers to set output power levels of transmtters, Base Statons (BSs) n the downlnk and UEs n the uplnk. A PC formula has been already agreed n a 3GPP meetng for the Physcal Uplnk Shared Channel (PUSCH) [2]. Fg. s based on an OLPC algorthm and CLPC adjustments can also be appled. The 3GPP specfcatons [3] defnes the settng of the UE transmt power P for PUSCH by the followng equaton P mn Pmax, P log M PL f ( ) msc () where P max s the maxmum UE transmt power, P s a parameter that has a cell specfc and nomnal part. It s measured n dbm/hz, expressng the power to be contaned n one Physcal Resource Block (PRB), M s the number of assgned PRBs to a certan user, s the cell-specfc pathloss compensaton factor that can be set to. and from.4 to. n steps of., PL s the downlnk path-loss measured n the UE, msc s a UE-specfc parameter (optonally cellspecfc), and f ( ) s a UE-specfc close-loop correcton value wth a relatve or absolute ncrease. The scope of PC s to defne the transmttng power n one PRB accordng to (), lettng the UE scale t to the assgned transmsson bandwdth (BW). Ths mples that ultmately t wll transmt wth a constant power n each assgned PRB, For ths reason, the term log M can be extracted from (). Fnally, removng the closed loop term, the Power Spectral Densty (PSD) formula results n (2), whch s referred to as the Fractonal Power Control (FPC) formula. PSD P PL dbm/hz (2) It s preferred to work wth the path gan nformaton whch s the lnear nverse of the path loss. Then, (2) s rewrtten as (3) n dbm PSD P PG dbm/hz (3) where PG s the path gan of the user to the servng BS. If =, a case referred to as no compensaton. All UEs wll transmt at full power whch results n hgh nterference level and poor cell edge performance. Wth =, a case referred to as full compensaton. The equaton reduces to tradtonal slow power control scheme where all UEs are receved at the same power resultng n poor spectral effcency. By lettng < <, one can acheve both good edge performance and hgh spectral effcency by lettng UEs wth good channel condton transmt at relatvely low power level to reduce the nterference. At the same tme, UEs wth bad channel condton are transmttng at relatvely hgh power level to acheve hgh spectral effcency. Regardng to one of the references [4], we wll use =.8 and P 8 dbm/prb whch acheve both good edge and cell throughput. Impact on the CINR Dstrbuton The Carrer to Interference plus Nose Rato (CINR) s one of the factors that determne the user throughput. Therefore, a dscusson of the mpact of the OLPC parameters on each UE experenced CINR would be very helpful for the operator. Let s defne the experenced CINR per user sd j E[ psd k ] pg k, j mw/hz (4) ks( j) where, sd j s the average nterference spectral densty perceved by a gven BS, s(j) denotes the users not served by BS j and allocated to transmt on the observed PRB, psd k s the power spectral densty for user k whch s not servng by the gven BS, pg k, j and the gven BS. psd pg, s( CINR = E sd s( ) n where n s the thermal nose. s the path gan between user k ) (5) Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 2
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons The second technque s Cell Interference Based Power Control C-IPC, whch proposed for each UE to have not less the mnmum reference CINR. In our work, we wll use the GI-PC as the second PC reference. The power spectral densty can be obtaned from (8) Fgure. PUSCH power control parameters broadcasted by BS towards the UEs Impact on the Cell and Edge throughput In EUTRAN LTE UL, the Modulaton and Codng Scheme MSC s chosen accordng to the state of CINR, hgher orders are used when ths s hgher. Equaton (6) shows how the user throughput s calculated for a gven user from ts experenced CINR and allocated bandwdth. [4] CINR C BWeff MBWPRB log 2 bps (6) Seff where BWeff s the bandwdth effcency Set to.72, s a correcton factor set to.68, M s the number of allocated PRBs, BW s the bandwdth of one PRB PRB Equal to 8 KHz, S eff s the CINR effcency at system level Set to.2 db. By takng one PRB to be compatble wth the fractonal frequency reuse whch wll be dscussed later, so there wll be dfference between our edge throughput and the reference edge throughput [4]. Equaton (7) s to calculate the cell throughput. T=E[C]*total number of PRBs at the system bps (7) where T s the cell throughput, E[C] s the average UEs throughput. Edge throughput s the lowest 5 % of Cumulatve Dstrbuton Functon (CDF) of the total cell throughput. B. Closed Loop Power Control: There are dfferent technques are used n CLPC because t does not have standardzaton. But the man dea of the closed loop s to start wth OLPC then the UEs also sends feedback to the BS, whch s then used to correct the user Transmtted T X power. There are two man technques used for CLPC, Generalzed Interference Based Power Control GI-PC, whch take n the consderaton the path loss to the servng BS, and the generated nterference from the UEs to the neghbour BS. PSD I PG PG dbm/hz (8) s where I s nterference power spectral densty lmt, t work as p n OLPC but the man dfference s that I s the power spectral densty per hertz but p s the total power contaned n one PRB, PGs s the path gan to the servng BS, PGI s the path gan to the nearest nterfered BS from the UE, s a parameter that affects the mpact of PGs on the T X PSD, s a parameter that affects the mpact of PG on the T X PSD. Impact on the CINR Dstrbuton The CINR can be easly obtaned same as OLPC but the man dfference wll be only n the PSD term. S I PG s (9) PG I I N where I s the average nterference spectral densty perceved by a gven BS and N s the thermal nose. For the cell and edge UEs throughput wll be the same as OLPC, other assumpton wll be at the Table 3. C. Frequency Reuses Schemes: There are three major technques used Hard Frequency Reuse (HFR), hard frequency reuse splts the system bandwdth nto a number of dstnct sub-bands accordng to a chosen reuse factor and lets neghborng cells transmt on dfferent sub bands see Fg. 2b. Fractonal Frequency Reuse (FFR), Fractonal frequency reuse [5] splts the gven bandwdth nto an nner and an outer part. The nner part s completely reused by all BSs, the outer part s dvded among the BSs wth a frequency reuse factor greater, as one seen n Fg. 2c. Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 22
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons TABLE SYSTEM MODEL DETAIL Smulaton case ISD meters BW MHz PLoss db Speed Km/h 5 2 3 TABLE 2 SYSTEM MODEL DETAIL Parameter Assumptons Cellular Layout Hexagonal grd, 9 cell stes, 3 sectors per ste (wrap around) Dstance-dependent path loss Penetraton Loss Antenna pattern(horzontal) (For 3-sector cell stes wth fxed antenna patterns) Shadowng modeled as a log-normal dstrbuton (SF) Total path loss Max UE Tx power Number of users n system Fgure 2. Dfferent frequency reuses technques Soft Frequency Reuse (SFR), soft frequency reuse [6][8][9], the overall bandwdth s shared by all base statons (reuse factor of one s appled), but for the transmsson on each sub-carrer the BSs are restrcted to a certan power bound see Fg. 2d. L=28. + 37.6*log(R) R n klometers 2 db 2 A ( ) mn 2, 2 7 PL=L+A(θ)+SF Mean =, standard devaton= 8dB L+A(θ)+SF 24 dbm *3*9=57 user db () where L s the path loss between BS and UE, A(θ) s the modeled antenna gan and SF s the shadowng III. PROPOSED SYSTEM MODEL In ths secton, the system model s dscussed; detals are shown n tables, 2. Followng the 3GPP gudelnes [], the cell smulaton layout consst of a wrap around Macro-cell scenaro reference case ; see Fg. 3. Composed by a grd of 9 stes wth 3 sectors each (9 BS wth 3 sectors, total cells are 9*3=57cells), each cell has user, the nter ste dstance s 5 meters and each sector s modeled by a hexagon The operatng bandwdth s dvded n 5 PRBs (48 PRB for users and 2 for sgnalng) wth a bandwdth of 8 KHz each. There s a Maxmal Rato Combnng (MRC) n the specfcatons, used to constructvely combne the multple receved sgnals n the antennas. It s modeled here as a constant gan of 3 db n the receved sgnal. The total path loss between an UE and a BS s modeled as n (). Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 Fgure 3. Wrap around Macro cell model 23
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons TABLE 3 PC PARAMETERS Parameter Value Unt Bandwdth effcency.72 bps/hz PRB bandwdth 8 KHz Max UE power 25 mw Number of PRBs per user - Outage 5 % Thermal nose level -74 dbm/hz Total number of 48+2 for PRBs sgnallng - Number of users per cell user MRC gan 3 db α.8 - P -8 dbm/hz I -57 dbm/hz β.7 - γ.3 - For the PC parameter we wll take the same assumpton as [4], except the PRB for each user wll be one PRB, all parameters are shown n Table 3. For the frequency reuse For HFR, we wll dvde the total used PRBs for the 3 sectors whch wll gve 6 PRBs for each sector. For FFR, we wll dvde the total PRBs to two groups each group s 24 PRBs, 24PRBs for the centre UEs (Ues, whch have path loss less than 2dB), and 24 PRBs s dstrbuted to the three sectors (8PRBs for each sector for the UEs whch have path loss more than 2dB). For SFR, we dvde the total PRBs to three [9] groups, the frst group nclude the UEs whch have path loss less than db, the second group nclude the UEs, whch have a path loss between db and 2dB and the last group nclude the UEs, whch have a path loss more than 2dB. IV. RESULTS AND ANALYSIS The mplementaton and smulatons are carred out usng a mult-cell rado network dynamc smulator mplemented n MATLAB to evaluate the PC wth dfferent FR schemes. The results show that all technques start from the lowest cell throughput and edge throughput and both of them ncrease to a certan pont, peak edge throughput observed when the frst user reaches the maxmum UE power lmtaton. Sudden decreasng appears n edge throughput due to nterference ncreasng regardng to the many UEs reach the maxmum power lmtaton whch leads to average PSD ncreasng, whch s responsble of edge throughput decreasng. We wll dvde the results to three man parts, valdaton results, OLPC wth dfferent FR schemes and CLPC wth dfferent FR schemes A. Valdaton results Fg. 4 shows a comparson between the obtaned results and that had been presented n [4] n the same operatonal condtons. It s shown that the obtaned results get more gans and have the same behavour of [4] takng n the consderaton that n [4] there are 6 PRB for each user,but n our case there are only PRB for each user to be compatble wth each FR scheme. B. OLPC wth dfferent FR schemes Fg. 5 llustrates dfferent schemes of FR. It s shown that by decreasng the nterference level by usng dfferent FR there wll be an ncreasng n the CINR. The obtaned results may be categorzed nto two man sectons. The frst one s the edge throughput and the other one s cell throughput.. Impact on edge throughput All FR schemes obtaned edge throughput gan over the ordnary OLPC. OL-HFR has become the hghest obtaned edge throughput, on the other hand OL-SFR s actng as the lowest edge throughput..8.6.4.2 OLPC [4] OL-HFR Present result - -5 5 5 2 OLPC as a valdaton curve vs OL-HFR CINR dstrbuton [db] Fgure 4. Shows there are CINR shft towards ncreasng wth HF reuse scheme Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 24
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons.8 OLPC [4] OL-HFR OL-FFR OL-SFR 5 5 4 5 OLPC [4] OL-HFR OL-FFR OL-SFR.6 3 5.4 2 5.2 5 - -5 5 5 2 OLPC wth dfferent FR schemes CINR [db] Fgure 5. CINR dstrbuton of OLPC wth. 8, P 8 dbm /Hz wth dfferent FR schemes, there are an ncreasng n CINR for all FR The results may be explaned as follows; OL-HFR: As a result of takng sxteen PRBs only for each cell, The nterference level s decreased by /3 compared wth ordnary OLPC; see Fg. 6. OL-FFR: Has a moderate edge throughput due to degradaton of nterference level by /3; see Fg. 6. OL-SFR: Has the lowest edge throughput due to the ncreasng of the nterference level when t s compared to the other FR schemes; see Fg. 6. 2. Impact on cell throughput Both OL-FFR and OL-SFR obtaned cell throughput gan over ordnary OLPC on the other hand CL-HFR has lower cell throughput than ordnary OLPC. The result may be explaned as follows; OL-HFR: Has the lowest cell throughput as the total number of PRBs s decreased to 6 PRBs only; see Fg. 6. OL-FFR: Has a good cell throughput regardng to decreasng the amount of nterference whch s generated from the edge UEs; see Fg. 6. OL-SFR has the hghest cell throughput because of decreasng the total amount of nterference for the cell; see Fg. 6. Fgure 6. Edge throughput vs. Cell throughput of OLPC wth. 8, P 8 dbm/hz and wth all FR schemes, there are edge throughput ncreasng for all FR over OLPC C. CLPC wth dfferent FR schemes Fg. 7 llustrates dfferent schemes of FR. It s shown that by decreasng the nterference level by usng dfferent FR there wll be an ncreasng n the CINR. The obtaned results may be categorzed nto two man sectons. The frst one s the edge throughput and the other one s cell throughput.. Impact on edge throughput All FR schemes obtaned edge throughput gan over the ordnary CLPC. CL-HFR has become the hghest obtaned edge throughput, on the other hand CL-SFR s actng as the lowest edge throughput. The results may be explaned as follows; 2 6 4 6 6 6 8 6 7.2 7 OLPC wth dfferent FR schemes cell throughput [bps] CL-HFR: The nterference level s decreased by /3 compared wth ordnary CLPC; see Fg. 8. CL-FFR: Has a moderate edge throughput due to degradaton of nterference level by /3; see Fg. 8. CL-SFR: Has the lowest edge throughput due to nterference level s hgher than the other two FR schemes; see Fg. 8. Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 25
ICDT 22 : The Seventh Internatonal Conference on Dgtal Telecommuncatons.8 CLPC [4] CL-HFR CL-FFR CL-SFR CL-FFR: Has a good cell throughput regardng to decreasng the amount of nterference whch s generated from the edge UEs; see Fg. 8. CL-SFR has the hghest cell throughput because of decreasng the total amount of nterference for the cell; see Fg. 8..6.4.2 I 57 - -5 5 5 2 CLPC wth dfferent FR schemes CINR [db] Fgure 7. CINR dstrbuton of CLPC wth. 7,. 3, dbm/hz wth dfferent FR schemes, there are an ncreasng n CINR for all FR 2. Impact on cell throughput Both CL-FFR and CL-SFR obtaned cell throughput gan over ordnary CLPC on the other hand CL-HFR has lower cell throughput than ordnary CLPC. The result may be explaned as follows; 5 5 4 5 3 5 2 5 5 CL-HFR: Has the lowest cell throughput as the total number of PRBs s decreased to 6 PRBs only; see Fg. 8. CLPC [4] CL-HFR CL-FFR CL-SFR 2 6 4 6 6 6 8 6 7.2 7 CLPC wth dfferent FR schemes cell throughput [bps] Fgure 8. Edge throughput vs. Cell throughput of CLPC wth.7,. 3, I 57 dbm/hz and wth all FR schemes, there are an ncreasng n all edge throughput V. CONCLUSION AND FUTURE WORK As the FR and PC functonaltes s a strong tool for cochannel nterference mtgaton, usng them crtcal ssues n cellular (OFDMA)/LTE networks. Both of OLPC & CLPC technques had been nvestgated. The novelty of the current work s presented va consderng both of FR schemes as well as the PC technques.the obtaned results shows gan n CINR for all FR schemes. The closed loop power control has shown more cell and edge throughput and system gan. Durng ths work PC technques wth dfferent FR schemes were analyzed by the means of a fxed bandwdth, balanced load and specfc boundres of PL for FR schemes. Future work could nvestgate the mpact of varable bandwdth and unbalanced load. An mportant contrbuton would be to fnd a mechansm to automatcally set the optmum boundres of PL for FR schemes and the ablty to swtch between dfferent FR schemes to obtan the best performance REFERENCES [] 3GPP TR 25.84, Physcal layer aspects for evolved unversal terrestral rado access (UTRA) (release 7), Tech. report, v7.., 26. [2] R-73224, Way forward on power control of PUSCH, 3GPP TSG- RAN WG 49-bs, 27. [3] 3GPP TS 36.23 V8.2., E-UTRA Physcal layer procedures, 28. [4] Nestor J. Quntero, Advanced Power Control for UTRAN LTE Uplnk, October 28. [5] Yong Soo Cho, Jaekwon Km, Won Young Yang, and Chung G. Kang, MIMO-OFDM wreless communcaton wth matlab, 2. [6] R-557, Soft Frequency Reuse Scheme for UTRAN LTE, Huawe.3GPP TSG RAN WG Meetng #4, Athens, Greece, May 25. [7] Carlos Ubeda Castellanos, Dmas Lopez Vlla, C Rosa, Klaus I Pedersen, F D Calabrese, Per-Henrk Mchaelsen, Jurgen Mchel, Performance of Uplnk Fractonal Power Control n UTRAN LTE, VTC Sprng 28 IEEE Vehcular Technology Conference (28) Publsher: Ieee, Pages: 257-252. [8] Ashley Mlls, Davd Lster, and Marna De Vos, Understandng Statc Inter-Cell Interference Coordnaton Mechansms n LTE, Journal of communcatons, vol. 6, no. 4, July 2. [9] Mathas Bohge, James Grossy, and Adam Wolsz, Optmal Power Maskng n Soft Frequency Reuse based OFDMA Networks, n Proc. of the European wreless conference 29 (EW 9), Aalborg, Denmark, May 29. Copyrght (c) IARIA, 22. ISBN: 978--628-93-9 26