PERFORMANCE OF THE LOCAL AVERAGING HANDOVER TECHNIQUE IN LONG TERM EVOLUTION NETWORKS

Transcription

1 Vol.6 (4) December 5 PERFORANCE OF THE LOCAL AVERAGING HANDOVER TECHNIQUE IN LONG TER EVOLUTION NETWORKS I.O. Elujde*, O.O. Olugbara**, P.A. Owolaw*** and T. Nepal**** Dept. of Informaton Technology, Durban Unversty of Technology, Durban 4, South Afrca Emal: *dasel@yahoo.com, **oludayoo@dut.ac.za, ****nepal@dut.ac.za *** Dept. of Electrcal Engneerng, angosuthu Unversty of Technology, Durban, South Afrca Emal: owolaw@mut.ac.za, p.owolaw@gmal.com Abstract: In ths paper, we nvestgate the performance of an alternatve receved sgnal flterng technque based on local averagng to mprove the qualty of handover decsons n Long Term Evoluton (LTE) networks. The focus of LTE-Advance (LTE-A) networks s to provde enhanced capacty and relablty of rado access as well as broadband demand for moble users. The necessty to mantan qualty of servce, especally for the delay senstve data servces and applcatons, has made moblty and handover decsons between the base statons n the LTE networks crtcal. Unfortunately, several handover decson algorthms n the LTE networks are based on the Reference Sgnal Receved Power (RSRP) obtaned as a lnear averagng over the reference sgnals. The crtcal challenge wth the lnear averagng technque s that the lmted reference sgnal avalable n the downlnk packet ntroduces an estmaton error. Ths estmaton error s a result of the effects of lnear averagng on propagaton loss components n elmnatng fast-fadng from the receved sgnals. oreover, prompt and precse handover decsons cannot be based on naccurate measurement. The standardzed LTE layer 3 flterng technque s appled to the local averaged layer sgnal to render t sutable for LTE handover decsons. The local averagng technque produces better handover than the lnear averagng technque n terms of the reduced number of handover falures, mproved hgh spectral effcency and ncreased throughput, especally for cell-edge users wth hgh speeds. The fndngs of ths study suggest that the local averagng technque enhances moblty performance of LTE-Advance networks. Key words: averagng, evoluton, flterng, handover, sgnal, network. INTRODUCTION The Long Term Evoluton (LTE) standard has evolved to the LTE-Advance n the Thrd Generaton Partnershp Project (3GPP) release []. The specfcaton of rado access networks was renewed to enhance the capablty and relablty of the networks. The capablty enhancement was acheved n the LTE networks because of the orthogonal frequency dvson multplexng technology employed n the rado nterface. The LTE rado nterface technology supports a transmsson protocol that uses both the Frequency Dvson Duplex (FDD) and the Tme Dvson Duplex (TDD) mechansms. The transmsson mechansms n downlnk and uplnk are based, respectvely, on the Orthogonal Frequency Dvson ultple Access (OFDA) and Sngle Carrer Frequency Dvson ultple Access (SC- FDA) controls. The OFDA allows for robustness of nter-symbol nterference whle enablng data and physcal layer sgnals to be concurrently multplexed. The smallest unt of resources transmtted n one OFDA symbol that corresponds to one subcarrer s called a resource element. A group of resource elements that corresponds to subcarrers n the frequency doman n one OFDA symbol s called a resource block. Some resource elements wthn a resource block are reserved for specal functons such as control sgnalng, system nformaton broadcast, cell search and synchronzaton, whle the remanng resource elements are used for data transmsson. The reference sgnals (RSs) whch are multplexed nto resource elements are used by a user equpment (UE) to determne the RSRP and Reference Sgnal Receve Qualty (RSRQ) []. The RSRP report from a UE s used to estmate the propagaton channel condton. Some other methods that could be appled to estmate the channel condton nclude the explotaton of correlaton propertes of the channel, the use of deductve knowledge of a parametrc model of channel, and blnd estmaton [3, 4]. The use of RSs for channel estmaton s the most common soluton because t s smple to mplement [3]. However, ths smplcty trades off spectral effcency for trackng varatons n the channel and reduces channel estmaton accuracy because of the lmted number of RSs avalable wthn each sub-frame. The lmted RSs are the prmary reason that the avalable RSs n adjacent subframes are exploted to yeld more accurate results [3, 5]. The use of the lnear averagng technque over the RSs n the dervaton of RSRP lmts the capablty of the RSRP to make fast, accurate handover decsons for UE, especally when a sudden attenuaton n the receved sgnal s experenced by the UE [6, 7]. Ths s because lnear averagng over RSs, necessary for removng the effects of fast-fadng, nterferes wth other propagaton loss components such as shadowng and path loss, and hampers the accuracy of the channel estmaton.

2 Vol.6 (4) December 5 3 Fgure : odel of an OFD transmtter Several varatons of the lnear averagng technque have been proposed n lterature n an attempt to strke a balance between accuracy and complexty [8]. An alternatve technque based on the local averagng can be used to solve the problem arsng from the removal of fast-fadng from the receved sgnal [7, 9, ] as well as assst n the protecton of the ntegrty of other propagaton loss components. It s partcularly germane to note that the accuracy of the estmaton made by a UE from the receved sgnal depends largely on the averagng technque employed and has a sgnfcant mpact on the ablty of the handover algorthm to make fast, relable decsons [, ]. Applcaton of local averagng technque n cellular networks was demonstrated n [3]. It should be noted that LTE networks belong to the famly of 3GPP technologes. oreover, layer (L) flterng s not restraned by the 3GPP standards [4]. The local averagng technque s therefore nvestgated n LTE networks to acheve fast and relable handover decsons for delay-senstve data servces, especally for a handover decson occurrng at the cell-edge n the presence of multple nterferng sgnals from neghborng cells. Ths paper reports the performance comparson of handover decsons made by usng local averagng (L flterng) and lnear averagng (L flterng) of RSRP. Performance s evaluated n terms of throughput, spectral effcency and average number of handover falures between the UE speeds of 3km/h up to km/h [5]. The remander of ths paper s succnctly organzed as follows: handover averagng technques are frst dscussed; the smulaton method and the parameters to compare the performances of the two averagng handover technques are thereafter presented; and fnally, results of the smulaton experments are presented. The paper concludes wth a bref statement of achevement.. HANDOVER AVERAGING TECHNIQUES A handover s a process of transferrng a UE call or a data sesson from one cell ste to another cell wthout dsconnectng the sesson. The tasks of a handover can be classfed appostely nto handover measurement, handover processng and handover decson. Herewth we dscuss these handover tasks.. Handover easurement The effects of nterference on the sgnal receved by a UE n a typcal wreless propagaton are classfed nto path loss, shadowng and fast-fadng [, 6]. The correspondng values of these losses, antenna gan and power transmtted from an enodeb wthn the operatng bandwdth are measured by the UE to facltate dynamc allocaton of network shared resources (E-UTRAN). The UE needs to provde a base staton wth measurement values of ts downlnk channel qualty, from ts cell as well as from neghborng cells, to facltate the selecton of an approprate cell to connect to the UE. The measurement of UE s necessary for moblty of a user wthn the E-UTRAN. The UE measurement s performed usng RSRP over cell-specfc RSs whch are multplexed nto the OFD resource elements and transmtted by some subcarrers. The RS s avalable to all UEs n a cell for determnng a phase reference demodulaton of downlnk control channels and for generatng the Channel State Informaton (CSI) feedback [3]. Channel estmaton s acheved through the transmtted OFD sgnal usng a flterng technque. Typcal mplementaton of an OFD transmtter s shown n Fgure. The transmtted seral data symbol s passed through a seral to parallel converter to generate L-dmensonal T parallel data block S [ [ S[, S[,..., S L [ ]. Each component of the parallel data stream s ndependently modulated, resultng n a complex vector T Y [ [ Y[, Y[,..., Y L [ ] used as an nput to an Inverse Fast Fourer Transpose (IFFT) system to generate tme doman complex samples usng Equaton : y m m [ YL[ exp(jl ) () L A conjugate operaton s performed at the recever usng an equvalent FFT operaton to obtan the frequency doman vector of the transmtted sgnal. If y(t) s the transmtted symbol at tme (t) where h(t) s a contnuous tme channel mpulse and n(t) s addtve nose, then the

3 4 Vol.6 (4) December 5 receved sgnal, x(t), n a multpath envronment s represented usng the receved dscrete tme OFD symbol x[ wth cyclc prefx, CP gven as: x[ y[ y x [ y [ y [ y CP [ h[ n[ [ hcp[ n [ k [ y CP ] To smply matters, Equaton can be wrtten further as follows: () x[ h[ n[ (3) x [ hcp[ n [ The FFT of matrx contanng the subcarrers wth varyng peak values produces a dagonal matrx [7, 8]. Ths mples that the matrx s equvalent to F H YF where F s a FFT matrx and Y s a dagonal matrx whose elements are gven, respectvely, by m F[ exp( L ) and m Ym[ yl[ exp( jl ). L The frequency doman representaton of the receved sgnal sample X[, after applyng the FFT, s gven by Equaton 4: X [ Y [ H[ N[ (4) X [ y [ H [ N [ The Channel Frequency Response (CFR) denoted by H can be expressed n terms of the Channel Impulse Response (CIR): h as H F. h [7].. Handover Processng The handover processng, also called the L flterng as defned by 3GPP, s performed at the LTE physcal layer [4]. The purpose of ths flterng s to remove the effects of fast-fadng n the sgnal receved by a UE. Lnear averagng and local averagng are two handover processng technques studed for ths purpose. Lnear averagng s the common technque for estmatng a channel over an RS. The channel estmaton s performed ether n the frequency doman or the tme doman [3, 8]. The estmate of the channel s then performed usng nterpolaton at several RS postons. For nstance, the decorrelaton of RS performed n the frequency doman s to determne the Channel Transfer Functon (CTF) as gven by Equaton 5: vˆ F h v ~ (5) Where s a value wthn the nterval (,..., ) ; s the number of the avalable RSs; F h s the same as the CFR; and v ~ s the whte nose vector. If a generc lnear flter D s used n the nterpolaton scheme for determnng an estmate of a channel over a subcarrer at ndex n then the CTF at subcarrer n can be wrtten as follows: n v vˆ Dˆ (6) The estmaton error of the nterpolated CTF of subcarrer n can be expressed as the dfference between the actual value and the estmated value as follows: v ~ n ( F DF ) h Dv ~ (7) The common lnear flters make use of technques such as Least-Squares (LS) and nmum ean-square Errors (SE) [8, 7, 9]. The LS s smple to mplement, but t cannot be appled drectly to LTE networks because of the ll-condtonng of the matrx nverse on the unmodulated subcarrers [3]. The SE produces a more accurate estmate than the LS; however, the SE s computatonally expensve because t requres second order characterstcs of the channel to perform the channel estmaton [7]. Local averagng, an alternatve technque to the lnear averagng technque, s performed as a convoluton of the exponental flter wth the downlnk receved sgnal [7]. Ths averagng technque s based on the local scatterng functon that estmates the power spectrum of the measured data usng an orthogonal wndow [9, ]. The ndvdual estmates of the spectra from the ndependent wndow data are aggregated by averagng to obtan a low varance estmate of the channel [, ]. If the CFR of the sampled spectra n tme and frequency domans s represented by H [ x, y], and assumng that the ndex of each tapped spectral at a specfc perod corresponds to w [ t, f ], the relatve sampled spectral ndces n tme and frequency domans are gven, respectvely, by x x y y and y' (,..., ) x' (,..., ). Where x s a tme value from the nterval of (,..., X ) and y s a frequency value from the nterval (,..., Y ). The estmate of the local scatterng functon at each ndex correspondng to a dscrete sample s gven by Equaton 8: Where N ( ) [, ] Qp wt w f H [ wt, w f ] (8) N p

4 Vol.6 (4) December 5 5 ( Q ) X ' X x Y H p [ w, w ] H. Q (9) t f ' Y y Where and N denote, respectvely, the total number of tapped spectral used n both the tme and frequency domans. The parameter Q p s the wndow functon equvalent to the exponental flter used for local averagng and N av s the averagng wndow sze. The parameter Q s determned as follows: p n Nav e Q p Nav.3 Handover Decson n,,,3,... () e The UE keeps track of the receved sgnal measurement from ts servng cell and neghborng cells n order to recognze the cell wth the best sgnal at the current poston. The report of ths measurement s sent to the servng enodeb. Sgnal attenuaton s experenced by the UE as t moves towards the target enodeb n the strongest nterferng cell from the servng enodeb. In other words, the sgnal receved from the servng enodeb gradually deterorates, whle that from the neghborng enodeb (target enodeb) gradually ncreases. At a partcular dstance from the servng enodeb, the receved sgnal from the servng enodeb goes below the handover threshold, a predefned value n the enodeb. The farther a UE moves away from the servng enodeb, the more the sgnal attenuates, whle there s a correspondng ncrease n the receved sgnal from the target enodeb. At a pont between the two enodebs, the receved sgnal from the servng enodeb becomes lower than that from the target enodeb. Fgure llustrates the concept of handover margn, the maxmum dfference between the values of the receved sgnals from two enodebs that can be tolerated before trggerng a handover decson. The regon beyond the handover margn where a handover decson occurs s called the handover regon. Handover margn s consdered n a handover decson before movng a UE to the target enodeb. Receved sgnal Servng enodeb Handover argn Target enodeb Handover Regon p 3. SIULATION The local flterng handover technque descrbed n the prevous secton was mplemented usng the system level smulator [5] as a new module n the LTE networks. The orgnal flterng technque mplemented n the smulator s based on lnear averagng, whle the local averagng technque was mplemented as a unque contrbuton of ths study. The lnear flterng algorthm mplemented n the smulator was to determne the utual Informaton Effectveness SINR (Sgnal to Interference plus Nose Rato) appng (IES) to serve as a baselne for comparson wth the mplemented local averagng flterng technque. Two handover decson algorthms were mplemented usng the two flterng technques to assess ther performances on the overall network. Fgure 3 shows a mcrocell network layout wth the hexagonal grd usng seven tr-sector stes (cell, cell and cell ) for the smulaton experments. The nter-ste dstance for each scenaro was chosen accordng to the recommendaton of ITU Rado communcaton (ITU-R) [6]. System bandwdth of Hz wth 5 resource blocks and GHz carrer frequency was used for the smulaton experments. cell cell cell nter ste dstance Fgure 3: Network layout 7-stes hexagonal grd The number of UEs at the commencement of the smulaton experments was kept constant. The UEs were unformly dstrbuted over the network coverage and ther drectons were randomly chosen from the range of o to 36 o C. Each UE moved wth a constant speed throughout the entre smulaton. The speed of UE was chosen from 3 km/h, 3 km/h and km/h dependng on the scenaro [5]. The channel estmaton of the sgnal receved at UE was dependent on the path loss, shadow fadng and fastfadng [7-9]. The shadow fadng wth a standard devaton of 8dB and mean was used for the smulaton. The detals of the smulaton parameters are provded n Table. UE Dstance travelled by UE from enodeb Fgure : Receved sgnal from two enodebs and handover margn

5 6 Vol.6 (4) December 5 Table : Smulaton Parameters PARAETERS ASSUPTION cell layout Hexagonal grd ( enodeb, 3 sectors per enodeb) carrer frequency GHz resource block (PRB) 5 system bandwdth Hz, 8kHz per PRB enodeb Tx power 46 dbm L3 samplng TTI L3 flter coeffcent 4 UE per enodeb UE nose fgure 9 db packet scheduler proportonal far path loss log (R n km) db shadow fadng standard devaton = 8dB correlaton mean = correlaton between enodeb=.5 fast fadng wnner channel model 4. RESULTS AND DISCUSSIONS The results of the smulaton experments performed n ths study are dscussed n ths secton. Each smulaton experment was performed for the duraton of 5 TTIs (Transmsson Tme Intervals) to ensure the relablty of results. The system performance s evaluated usng the metrcs of throughput, spectral effcency and average number of handover falures for each of the L flterng technques. Performance metrcs have been selected to evaluate system and moblty-related performances [3, 3]. The throughput, the total number of the transmtted data packets per second, s measured n unts of bts per second (bps) [3, 3]. The spectral effcency whch ndcates the amount of spectrum used s the net UE data bt rate transmtted over the operatng bandwdth and s measured n bts per second per hertz (bps/hz) [9]. The values used for the averagng wndow N av n the local averagng flterng technques are, respectvely, 5, 6 and 8.5 for the correspondng UE operatng at standard speeds of 3, 3 and km/h. In Fgure 4, the emprcal Cumulatve Dstrbuton Functon (CDF) of the average UEs spectral effcency at 3km/h s presented. The emprcal CDF gves a far estmate of the UEs CDF and a consstent estmate of the real CDF at any gven pont [4]. We have observed that a handover algorthm based on the local averagng technque s slghtly more spectral effcent than the lnear averagng technque n terms of the rate of nformaton transmtted n number of bts per channel. There s no remarkable dfference n the spectral effcency for the th to 3th percentle, but the average user spectral effcency gradually ncreases from the 4th percentle to about the 95th percentle. Results ndcate that the capacty obtaned wthn the cell s hgher for average users and peak users when the local averagng flterng technque s used at ths speed. Emprcal CDF LTE Lnear averagng LTE Local averagng average UE spectral effcency [bps/hz] Fgure 4: Emprcal CDF of average UE spectral effcency at 3 km/h Fgure 5 presents the results obtaned when UEs are movng at 3 km/h n the smulated envronment. The emprcal cumulatve dstrbuton functon (emprcal CDF) shows that the probablty of the average UE spectral effcency s hgher when the local averagng technque was used for a handover decson, suggestng that the number of bts transported wthn the bandwdth at ths speed s hgher for the local averagng technque than for the lnear averagng technque. It s observed from ths result that there s a sgnfcant dfference between lnear averagng and local averagng n terms of the amount of nformaton transmtted by an average user at the th percentle to the 9th percentle. The local averagng technque produces hgher average user spectral effcency n bts per second per hertz than the lnear averagng technque. Emprcal CDF LTE Lnear averagng LTE Local averagng average UE spectral effcency [bps/hz] Fgure 5: Emprcal CDF of average UE spectral effcency at 3 km/h The emprcal CDF n Fgure 6 shows the average user spectral effcency n bts per second per hertz (bps/hz) when the UE speed s km/h. The results suggest that the lmted frequency spectrum s more utlzed when the local averagng technque s employed than when the lnear averagng technque s used, meanng that the average number of users accommodated to transmt call smultaneously over the lmted spectrum s hgher for the local averagng technque, although at about the 95th percentle there s only a slght dfference between the performances of the two averagng technques studed. However, there s a clear ndcaton of the mpact of dfferences n the averagng technque on the spectral effcency wthn a cell from the th percentle to about the 9th percentle.

6 Vol.6 (4) December 5 7 requres a hghly accurate flterng technque to keep track of the channel condtons. Emprcal CDF LTE Lnear averagng LTE Local averagng average UE spectral effcency [bps/hz] Throughput (bps) lnear averagng local averagng Fgure 6: Emprcal CDF of average UE spectral effcency at km/h Fgure 7 shows the results of the peak throughputs based on lnear averagng and local averagng technques. It can be observed from Fgure 7 that the effects of these flterng technques are not notceably dstngushable at a relatvely low speed of about 3 km/h. However, the local averagng technque acheves a better performance n terms of the peak throughput wthn the cells as the speed ncreases. It can be determned from Fgure 7 that whle the effect of the average multple ndependent spectra used by the local averagng technque s not clearly vsble at low speeds, t gves a better estmate of the channel qualty that s achevable by a UE as the speed ncreases. The mproved throughput experenced at hgher speeds when the local averagng technque s employed s due to the accuracy of the channel estmate whch nfluences the choce of CS and ncreases the data rate acheved by the UE. Throughput (bps) lnear averagng local averagng UE speed (kmph) Fgure 7: Peak user throughput at dfferent UE speeds The results of the smulaton experment for the average throughput experenced by a user are shown n Fgure 8. The performances of the two flterng technques are almost the same for the throughput experenced by the UEs. However, the average user throughput experenced when the local averagng technque was employed s slghtly hgher than that of the lnear averagng technque at hgher speeds. Ths s because at low speeds, the rate of change of the rado channel condton experenced by a user s very low, renderng the estmaton error of both flterng technques neglgbly small. At hgher speeds, however, the rado channel changes at a faster rate and UE speed (kmph) Fgure 8: Average user throughput at dfferent UE speeds Fgure 9 presents the results of the cell-edge user throughput at dfferent UE speeds. The cell-edge user throughput performance wth the local averagng technque s slghtly better than wth the lnear averagng technque at hgher user speeds. Although the cell-edge user throughput for the lnear averagng technque s not as hgh as that of the local averagng technque at low user speeds, the rate of change s not as remarkable as n the local averagng technque. However, the rate of change for the cell-edge throughput based on the local averagng technque s remarkably better than that of the lnear averagng technque at hgher speeds. Ths result translates to the perceved QoS experenced by the celledge users as the speed ncreases. The low speed users mght experence a sharp change n the QoS when the local averagng technque s used, though ths mght not be the case for a UE that employs the lnear flterng technque. However, the experence s reversed n the case of a UE at hgher speeds. Throughput (bps) lnear averagng local averagng 3 3 UE speed (kmph) Fgure 9: Cell-edge user throughput at dfferent UE speeds Fgure shows the average number of handover falures per UE speed. When the speed s as low as 3 km/h, the rate of handover falures obtaned s low for both of the handover flterng technques. The rate of handover falures due to the lnear averagng technque s as low as less than.5%. The average number of handover falures observed s also remarkably low for the local averagng technque, wth a value less than %.

7 8 Vol.6 (4) December 5 Average number of handover falure Lnear averagng local averagng local+l3 flterng 3 3 UE speed (kmph) Fgure : Effect of UE speeds on average number of handover falures As expected, the average number of handover falures ncreases as the user speed ncreases. At hgher speeds, the dfference n performance of both handover flterng technques s not partcularly sgnfcant. However, the effect of L3 flterng on handover falures becomes apparent at a hgher speed because L3 algorthms flter output used for trggerng a handover decson. In addton, ths reduces the L estmaton error, whch becomes hgher as user speed ncreases because of the hgh uncorrelated nature of the tme-varyng channel between the UEs and the base statons. 5. CONCLUSION Ths paper reports the performance of the local averagng technque n LTE networks. The evaluaton metrcs of throughput, spectral effcency and average number of handover falures establsh comparson between the local averagng handover technque and the lnear averagng handover technque usng UEs at varous speeds. Performance analyss shows the effect of each handover flterng technque on achevable capacty wthn the system n terms of spectral effcency, user throughput and moblty based on the average number of handover falures. The spectral effcency for pedestran speed (3 km/h) UEs for the local averagng technque when compared to the lnear averagng technque produces, respectvely, an ncrease of 9.%,.8% and 5.% for cell-edge, average and peak users. From the results obtaned at a UE speed of 3 km/h, the comparson between the lnear averagng technque and the local averagng technque shows, respectvely, ncreased capactes of about 3.6%, 37.9% and 5.3% for celledge, average and peak users. The spectral effcency at a hgher speed of km/h produces, respectvely, 5.%, 68.7% and 4.8% ncreased capactes for cell-edge, average and peak users. The system throughput for celledge users shows, respectvely, a 44.8%,.7% and 4.8% mprovement at UE speeds of 3 km/h, 3 km/h and km/h when the local averagng flterng was employed. The peak user throughput for the lnear averagng technque s 4.% better than that of the local averagng technque. However, the local averagng technque shows, respectvely, better performances of about 3.% and 7.4% at the UE speeds of 3 km/h and km/h. The results ultmately obtaned from the comparson of the average number of handover falures between the two L flterng technques show a sgnfcant reducton n the average number of handover falures of about 8.9% for pedestran users at the speed of 3 km/h usng the local averagng technque. The results at the UE speeds of 3 km/h and km/h show, respectvely, reductons of about.5% and 4.6% n average number of handover falures of the local averagng flterng technque. The applcaton of the L3 flterng n the local averagng technque further mproves performance by 6.9%, 8.6% and.8% at the UE speeds of 3 km/h, 3 km/h and km/h, respectvely. The results of ths study reveal that both handover flterng technques nvestgated are sutable for makng handover decsons n LTE networks. However, the local averagng technque could ensure the provsonng of hgher Qualty of Servce (QoS) on LTE networks because of the reduced average number of handover falures and mproved cell capacty as reflected by the spectral effcency. As mantanng QoS s partcularly germane, dverse sophstcated technques have been used to maxmze the performance of networks for achevng hgh user throughput. The dstrbuton of user throughput s a clear ndcator of QoS. In codcl, t shows data rates experenced by users at dfferent locatons wthn the cell: the 95% user throughput s consdered a peak throughput; the mean user throughput s consdered a typcal data rate achevable wthn the coverage area of the networks; whle the 5% user throughput s termed cell-edge user throughput. The results of user throughput at dfferent speeds as a result of applyng two handover flterng technques have been presented n ths paper. REFERENCES 5. [] UTS Forum: oble Broadband Evoluton: the roadmap from HSPA to LTE, whte paper from the UTS Forum, UK, February 9. [] 3GPP: 3rd Generaton Partnershp Project (3GPP); Techncal Specfcaton Group Rado Access Network; Evolved Unversal Terrestral Rado Access (E-UTRA); Physcal layer; easurements (Release 9), 3GPP TS 36.4 V9.. (-3),. [3] F. Tomats and S. Sesa: LTE - The UTS Long Term Evoluton from Theory to Practce, John Wley, UK, frst edton, chapter 7, pp. 5-64, 9. [4] D.T.. Slock: Sgnal processng challenges for wreless communcatons, n st Internatonal Symposum on Control, Communcaton & Sgnal Processng, pp , 4.

8 Vol.6 (4) December 5 9 [5] F. Khan: LTE for 4G moble broadband: ar nterface technologes and performance, Cambrdge Unversty Press, USA, frst edton, 9. [6] O. Grmlund and B. Gudmundson: Handoff strateges n mcrocellular systems, Proceedng of the 4st IEEE Vehcular Technology Conference, St. Lous, O, pp. 55-5, ay 99. [7] B.L. ark and A.E. Leu: Local averagng for fast handoffs n cellular networks, IEEE Transactons on Wreless Communcaton, Vol. 6 No. 3, pp , arch 7. [8] X. Da, W. Zhang, J. Xu, J.E. tchell and Y. Yang: Kalman nterpolaton flter for channel estmaton of LTE downlnk n hgh-moblty envronments, EURASIP Journal of Wreless Communcaton & Networkng, Vol. No., pp. -4,. [9] J.. Holtzman and A. Sampath: Adaptve averagng methodology for handoffs n cellular systems, IEEE Transactons on Vehcular Technology, Vol. 44 No., pp , February 995. [] R. Vjayan and J.. Holtzman: A model for analyzng handoff algorthms [cellular rado], IEEE Transacton on Vehcular Technology, Vol. 4, pp , August 993. [] A.H.S. Tamlselvan, and K. anvannan: Performance analyss of fast handoff algorthms usng Local averagng technque n cellular networks, Proceedng of the st Internatonal Conference of Dstrbuted Framework & Applcatons, Penang, pp , October 8. [3] 3GPP: 3rd Generaton Partnershp Project; Techncal Specfcaton Group Rado Access Network; Evolved Unversal Terrestral Rado Access (E-UTRA); Requrements for support of rado resource management (Release 9), 3GPP TS V9.. (9-), 9. [4]. Anas, F.D. Calabrese, P.E. Ostlng, K.I. Pedersen and P.E. ogensen: Performance analyss of handover measurements and layer 3 flterng for UTRAN LTE, Proceedng of the IEEE 8th Internatonal Symposum on Personal, Indoor & oble Rado Communcatons, Athens, pp. -5, September 7. [5] 3GPP: 3rd Generaton Partnershp Project; Techncal Specfcaton Group Rado Access Network; Requrements for Evolved UTRA (E- UTRA) and Evolved UTRAN (E-UTRAN) (Release 8), 3GPP TR 5.93 V8.. (8- ), 8. [6] G. Roche, A.A. Glazunov and B. Allen: LTEadvanced and Next Generaton Wreless Networks: Channel odellng and Propagaton, John Wley, Chchester U.K., frst edton,. [7]. Gudmundson: Correlaton model for shadow fadng n moble rado systems, IET Electroncs Letters, Vol. 7, pp , 99. [8] J.J. Beek, O. Edfors,. Sandell, S.K. Wlson and P. Ola Borjesson: On channel estmaton n OFD systems, Proceedng of the IEEE 45th Vehcular Technology Conference, Chcago, IL, pp vol., July 995. [9] A. Kalakech, L. Brunel,. Berbneau and D otter: A novel OFD power based estmaton for dynamc channel trackng n downlnk LTE, Proceedng of the IEEE 76th Vehcular Technology Conference, Quebec Cty, pp. -5, September. [] G. atz: On non-wssus wreless fadng channels, IEEE Transacton on Wreless Communcatons, Vol. 4 No. 5, pp , November 5. [] D. J. Thomson: Spectrum estmaton and harmonc analyss, Proceedng of the IEEE, Vol. 7 No. 9, pp , 98. [] D. Percval and A. Walden: Spectral analyss for physcal applcatons: multtaper and conventonal unvarate technques, Cambrdge Unv. Press, New York, 993. [3] H. Suzuk: A statstcal model for urban rado propagaton, IEEE Transacton on Communcatons, Vol. 5, pp , 977. [4]. Anas, F.D. Calabrese, P.E. ogensen, C. Rosa, and K.I. Pedersen: Performance evaluaton of receved sgnal strength based hard handover for UTRAN LTE, Proceedng of the IEEE 65th Vehcular Technology Conference, Dubln, pp. 46-5, Aprl 7. [5] J.C. Ikuno and. Taranetz: Venna LTE smulators, System Level Smulators verson.6r885, Venna Unversty of Technology, Austra,. [6] ITU: Gudelnes for evaluaton of rado nterface technologes for IT-Advanced, ITU-R.35-, 9.

9 Vol.6 (4) December 5 [7] 3GPP: 3rd Generaton Partnershp Project; Techncal Specfcaton Group Rado Access Network; Physcal layer aspects for evolved Unversal Terrestral Rado Access (UTRA) (Release 7), 3GPP TR 5.84 V7.., 6. [8] H. Claussen: Effcent modellng of channel maps wth correlated shadow fadng n moble rado systems, Proceedng of the IEEE 6th Internatonal Symposum on Personal, Indoor & oble Rado Communcatons, Berln, pp. 5-56, September 5. [9] L. Hentlä, P. Kyöst,. Käske,. Narandzc and. Alatossava: ATLAB mplementaton of the WINNER phase II channel model, [Onlne]. Avalable: org/phase model. html7, 7. [3] J. Gora, A. Gouraud, P. arsch, N. Gresset,. Karray, R. Irmer, and T. Svensson: Scenaros, key performance ndcators and evaluaton methodology for advanced cellular Systems, Advanced Rado Interface Technologes for 4G Systems, EU FP7 INFSO-ICT-473 ARTIST4G consortum,. [3] ETSI: LTE; Evolved Unversal Terrestral Rado Access (E-UTRA); Rado Resource Control (RRC); Protocol specfcaton (3GPP TS verson.5. Release ), ETSI TS V.5. (-3),. [3] H.. Raml, K. Sandrasegaran, R. Basukala and L. Wu: odelng and smulaton of packet schedulng n the downlnk long term evoluton system, Proceedng of the 5th Asa-Pacfc Conference on Communcatons, Shangha, pp. 68-7, October 9.