Optimal Decoders for 4G LTE Communication

Transcription

1 Lecture Notes on Informaton heory Vol. 1, No. 4, December 013 Optmal Decoders for 4G LE Communcaton Syed Al Irtaa, Aamr abb, and Qamar-ul-Islam Insttute of Space echnology, Islamabad-44000, Pakstan Emal: {syed.al, aamr.habb, Abstract Multple Input Multple Output (MIMO) technques are used to reale practcal hgh data rate systems whch lad the foundaton of Long erm Evoluton (LE). Varous transmsson technques lke ransmt Dversty (D), Open Loop Spatal Multpleng (OLSM) and Closed Loop Spatal Multpleng (CLSM) are deployed n the realm of MIMO. he spectral effcency s mproved wth the help Orthogonal Frequency Dvson Multpleng (OFDM). In ths paper we wll focus on CLSM, to evaluate ts performance wth the help of Zero Forcng (), Mnmum Mean Sqaure Error (), SoftSphere Decoder (SSD), SSD K-Best (SSDKB) and SIC recevers to fnd the optmal decoder n LE envronment. he SSD, SSD-KB and SIC uses based equalers. he channel envronment used are Addtve Whte Gaussan (AWGN), Vehcular A (VehA), Vehcular B (VehB) and an outdoor Pedestran (Ped B) channel model. A Least Square (LS) estmated feedback obtaned by the averagng of two channel nstances s used to mprove BLER n the case of fadng channels. MIMO confguraton and user speed. to facltate hgher data rate t provdes a fleble spectrum management and supports allocaton of multple bandwdth slots to same user on demand. I also supports multple antenna confguraton schemes from 44 MIMO to 11 SISO. dfferent transmsson schemes lke multpleng and transmt dversty to acheve hgh data rates nad provdes aupport for users movng upto a speed of 500Km/hr (310 M/hr) dependng upon the user terran. Multple decodng algorthms are avalable n LE communcaton network to reduce the bt error probablty. In ths paper we manly focus on Zero Forcng (), Mnmum Mean Sqaure Error (), SoftSphere Decoder (SSD), SSD K-Best (SSDKB) and SIC recevers to fnd the optmal decoder n LE envronment. he SSD, SSD-KB and SIC uses based equalers. he channel envronment used are Addtve Whte Gaussan (AWGN), Vehcular A (VehA), Vehcular B (VehB) and an outdoor Pedestran (Ped B) channel model. he smulaton results acheved meet ndustral standards wth the help of lnk level LE smulator [1] complant wth the parameters specfed by the 3GPP workng group. In the followng paper Secton II descrbes the channel model and recever algorthm. In Secton III, the CLSM mode of transmsson s eplaned. he Secton IV eplans outcome of these smulatons and observatons. Conclusons are gven n Secton V. Inde erms VehB, SSD, SIC, CLSM, LE, LS. I. INRODUCION Wreless communcatons contnue to strve for hgher data rates and a better lnk relablty n order to provde more advanced servces on the go. he use of multple antennas at both the transmtter and recever sde,.e., multple-nput multple-output (MIMO) communcatons, s one of the most promsng technologes to fulfll these demands. Indeed, MIMO systems are capable of achevng ncreased data rates and an mproved lnk relablty compared to sngle-antenna systems wthout the ad of addtonal bandwdth or transmt power. hese mprovements, however, requre the use of more computatonally ntensve data detecton algorthms at the recever sde. In partcular, optmum data detecton can easly become comple. Conventonal sub-optmum detecton technques have a low computatonal cost but ther performance s n general less sgnfcant to that of optmum data detecton. hus, there s a strong demand for computatonally effcent data detecton algorthms that are able to reduce ths performance gap. One of the promsng technologes to provde hgh data rate at hgh speeds whle mantanng the specfed Qualty of Servce (Qos) s Long erm Evoluton (LE). LE provdes a mamum downloadng data rate of 99.6Mbts/s and an uploadng data rate of 75.4Mbts/s constraned by the II. he proposed MIMO [] system model consstng of N transmt antennas and M R receve antennas, defned by the followng Equaton (1). y = R s the channel coeffcent matr of the dmensons M R N defnng the channel gan epected values = [ 1 M ] s the nose. s assumed to be R (..d) Zero Mean Crcularly Symmetrc Comple Manuscrpt receved June 10, 013; revsed August 9, Engneerng and echnology Publshng do: /lnt (1) h1, h1, N h1,1 h1, h, N h,1 () M,N = R h h hm, N R M R,1 M R, where y = [ y1 y ym ] s the receved vector, and CANNEL MODEL 170

2 Lecture Notes on Informaton heory Vol. 1, No. 4, December 013 Gaussan (ZMCSCG). he channel s defned by the channel delay profle. he nput s dvded nto dfferent streams of data wth the help of spatal demultpleer as n Fg. 1. he streams are than processed by the turbo decoder to provde communcaton at low values of SNR. IFF s used to provde computatonal effcency and cyclc pref s added to mantan synchronaton.he streams are passed through the nter-leaver after the channel codng s appled. he nter-leaver processes the nput such that the consecutve bts are placed far apart to avod burst error due to fadng. he modulaton scheme s than appled whch n ths case s 16-QAM wth an effectve codng rate of he modulated data s passed through the seral to parallel converter. On recepton data s processed wth the decoder. Fgure 1. MIMO transmsson scheme Recever Algorthm A bref descrpton of the recevers s gven below: Recever Zero-Forcng () detecton s the smplest and effectve technque for retrevng multple transmtted data streams at the recever wth very lttle complety. he probablty densty functon (PDF) for the sgnal-to nose-plus-nterference rato (SINR) at the output of a ero forcng () detector n a flat fadng channel was derved n [3], [4]. he ero-forcng () technque s used to nullfy the nterference wth the help of followng weght matr: W = 1 where. denotes the ermtan transpose operaton. In other words, t nverts the effect of channel as ~ = W y ; = ~ 1 where = W = (3) (4). Note that the error performance s drectly proportonal to the power of ~. (.e., ~ ). he post-detecton can be calculated usng SVD as Snce ~ 1 = ( ) 1 = V. V V U (5) (6) 1. U (7) = V Q untary matr s gven as = Q Q = = for Q, the epected value of the nose power a 1 =. E E U Recever Multple antennas offer sgnfcant performance mprovements n wreless communcaton systems by enablng communcatons by mnmng the error at hgher data rates. Lnear recevers lke mnmum-mean-squared-error () recever are a practcal soluton to provde lower complety and hgher data rates wth the ad of multpleng technques whch n our work s spatal multpleng. he recever s partcularly mportant as t optmally trade off strengthenng the energy of the desred sgnal of nterest and cancelng unwanted nterference by usng ts receve degrees of freedom (DOF) such that the sgnal-to nterference-and-nose rato (SINR) s mamed. In [5], multpleng at recever sde s used for spatal dversty to ncrease the desred sgnal power, whle n [6], multpleng at recever sde s used to cancel nterference from the strongest nterferer nodes. In [7], recevers are used and the average spectral effcency, a per-lnk performance measure, was obtaned n the large antenna regme. In [8-10], by usng sub-optmal and lnear recevers, the results of transmsson capacty were shown to scale lnearly wth the number of receve antennas. he post-detecton sgnal-to-nterference plus nose rato (SINR) can be mamed by usng the crtera, the weght matr s used whch gven as W = ( I) 1 (8) For recever to perform effcently, the statstcal nformaton of nose s requred. he th row vector w, of the weght matr n Equaton 8 s obtaned by solvng the optmaton equaton gven below: W tr, =arg mn wh N w=( w,..., w ) 1 N E wh w j=1, j Usng the weght n Equaton 8, we obtan the followng relatonshp: ~ = ( 1 1 = E tr(. U U. ) = tr(. 1 U E U. 1 ) tr 1 1 = (.. ) = (. ) = N =1 U U = W I) 1 y E (9) (10) 013 Engneerng and echnology Publshng 171

3 Lecture Notes on Informaton heory Vol. 1, No. 4, December 013 =~ ~ ~ = (( I ) 1 ). Usng where Sngular Value Decomposton (SVD), the post detecton nose power s gven by the Equaton 1. E = E. N = =1 1 1 U (11) For a recever t s preferable to have a hgh densty of sngle-stream transmssons than a low densty of mult-stream transmssons. hs s because n detecton, the nterference powers from the strongest nterferers source remanng after nterference-cancellaton are weaker for sngle stream transmsson than mult-stream transmsson. Soft Sphere Decoder SSD gves the ML soluton wth soft outputs. hese ML symbols are chosen from a reduced set of vectors wthn the radus of a gven sphere rather than a complete vector length. he radus of the sphere s adjusted such that there ests only one ML symbol wthn the gven radus. SSD provdes sub optmal ML soluton [11] wth reduced complety provded s used to estmate the channel. he Soft Sphere Decoder (SSD) soluton s gven by the followng equaton. Fgure 3. SIC recever. III. RANSMISSION MODELS MIMO mprove the spatal and multpleng gans by the use of dversty and spatal multpleng [1]. he methods used to enhance the dversty and multpleng gans s CLS Closed Loop Spatal Multpleng Independent data streams are transmtted from the N transmt antennas n CLSM Fg. 4. In CLSM essental amount of CSI s used as feedback whch enables us to acheve hgh throughput wth lower BLER. arg mn y = arg mn( ˆ ) ( ˆ) (1) where ( ) denotes the transpose of matr. Equaton 1 gves the unconstraned soluton of the real tme system. hs means that the ML soluton can be determned by the term ( ˆ ) ( ˆ ). No ML value ests outsde the sphere because there ML value s greater than those whch ests nde the sphere hence makng a unque detecton as n Fg.. Fgure 4. Block dagram of a MIMO transmsson usng CLSM. IV. SIMULAION RESULS AND DISCUSSION ABLE I. Parameters Recevers Channel User Speed Fadng ype Retransmsson Algo. No of Retransmssons Soft Demapper Modulaton Feedback Estmaton Feedback Bts Resource Blocks Fgure. Illustraton of the sphere n sphere decodng. K- Best Soft Sphere Decoder he K-Best SSD s a varant of SSD, and performs ts operaton on K best selected optons unlke the SSD whch consders only one pont. Successve Interference Canceller Decoder SIC recever s a collecton of lnear recever banks whch successvely cancels the nterference whch n ths case are recevers, as shown n the Fg. 3. M. 013 Engneerng and echnology Publshng LE SIMULAION PARAMEERS. Values,, SSD, SSDKB, SIC Veh A, Veh B, PedB, AWGN 30 Km/h, 10 Km/h and 3Km/h Block Fadng ARQ 03 Ma Log Map 16 QAM, CQI 9 Least Square In ths paper, ybrd Automatc Repeat Request (ARQ) s set to a mamum value of 03 to provde retransmsson n the case of fadng.e. block fadng n ths scenaro. Soft decsons are made usng the ma log map crteron for lower probablty of error. VehA and VehB channels are consdered for observng the LE lnk 17

4 Lecture Notes on Informaton heory Vol. 1, No. 4, December 013 performer n terms of throughput and BLER whle SSD-KB s provdng a sub optmal output. behavor. he feedback for supportng CLSM transmsson mode s obtaned by channel averagng of two channel realatons. A complete detal of the parameters used n the smulatons are gven by the able I. In case of AWGN channel, from Fg. 5 and 6 t can be seen that at hgher values of SNR all the recevers are performng equally good gvng almost the same throughput and BLER. In case of lower SNR, n AWGN channel SIC recever gves a better output as compared to all other recevers. Fgure 8. Recevers BLER n VehA channel usng CLSM. Fgure 5. Recevers throughput n AWGN channel usng CLSM. Fgure 9. Recevers throughput n VehB channel usng CLSM. Fgure 6. Recevers BLER n AWGN channel usng CLSM. Fgure 10. Recevers BLER n VehB channel usng CLSM. Fgure 7. Recever s throughput n VehA channel usng CLSM. In case of VehA channel, from Fg. 7 & 8 t can be seen that at hgher values of SNR, SIC recever s gvng the best out put n terms of throughput and BLER whle SSD-KB s provdng a sub optmal output. For the lower values of SNR, SIC s the best performer among all the recevers whle SSD s the second best. In case of VehB channel, from Fg. 9 and 10 t can be seen that at hgher values of SNR, SIC recever s the best 013 Engneerng and echnology Publshng Fgure 11. Recevers throughput n PedB channel usng CLSM. For the lower values of SNR, SIC s the best performer among all the recevers whle SSD s the second best.he BLER n ths case s worst among all the channel condtons and the recepton of correct data s hardly epected. In case of outdoor pedestran channel model Ped B, from Fg. 11 and 1 the performance of SIC s no 173

5 Lecture Notes on Informaton heory Vol. 1, No. 4, December 013 [6] dfferent as n the case of VehA and VehB channel. SSD and SSDKB performs almost same at the hgher values of SNR. At the lower values of SNR, the verson of SIC recever s performng better than the 44 versons of SSD, SSDKB, and recevers n terms of throughput and SNR. [7] [8] [9] [10] [11] Fgure 1. Recevers BLER n PedB channel usng CLSM. V. [1] CONCLUSIONS [13] In order to acheve hgher through put [13] n LE, SIC recever must be used n all channel models. Consderng the performance/complety trade off SSD and SSDKB recevers provde a reasonable output n terms of throughput and BLER as compared wth the SIC recever. hs performance/complety trade-off makes SSD and and ts varant SSDKB as the optmal recevers. A carefully desgned mechansm s needed to select the optmal recever accordng to the throughput and BLER requrements of the user keepng n vew the performance/complety trade-off n case of both hgh and low values of SNR. here s a great room for mprovement n terms of throughput and BLER wth the help of CLSM. hs can be mproved by ncreasng the number of plot channels or by ncreasng the number of bts per plot channel provdng the feedback whle conservng the communcaton standards specfed by the 3GPP and IU- to get the advantages of CLSM. Al Irtaa was born n Rawalpnd, Pakstan. e s wth the Insttute of Space echnology for the Masters degree n Wreless Communcatons and now servng as a Lecturer snce May 013. e dd hs Bachelors from COMSAS Unversty of Scence and echnology n 009. e has worked wth ZONG n Network Operatons Department. s area of nterests nclude Software Defned Rado and Advanced Wreless Communcatons. Aamr abb was born n Rawalpnd, Pakstan. e dd hs Masters n Moble and Satellte Communcatons from the Unversty of Surrey UK n 004 and another from Center for Advanced Studes n Engneerng, Islamabad n Computer Engneerng n 006. Fnshed hs Doctoral thess sponsored by gher Educaton Commsson, Pakstan n collaboraton wth Austran Echange servce (OeAD) Austra n Electrcal Engneerng. e s workng at the Insttute of Space echnology, Islamabad snce June 000. Research Interests nclude Advanced Moble Communcatons, Space me Processng Algorthms, Antenna Selecton, Algorthms for MIMO Communcatons, WMAX echnologes and Performance, Antenna Selecton Methods for LE. REFERENCES [1] [] [3] [4] [5] C. Mehlfuhrer, M. Wrulch, J. C. Ikuno, D. Bosanska, and M. Rupp, Smulatng the long term evoluton physcal layer, n Proc. 17th European Sgnal Processng Conference, Glasgow Scotland, Aug 009. Y. S. Cho, J. Km, W. Y. Yang, and C. G. Kang, MIMO-OFDM Wreless Communcatons wth Matlab, John Wley & Sons (Asa) Pte Ltd, 010. D. Gore, R. W. eath, and A. Paulraj, On the performance of the ero forcng recever n presence of transmt correlaton, n Proc. IEEE Int. Symp. Inform. heory, 00, pp P. L, D. Paul, R. Narasmhan, and J. Coff, On the dstrbuton of snr of the mmse mmo recever and performance analyss, IEEE rans. Inform. heory, vol. 5, no. 1, pp , Jan 006. A. M. unter, J. G. Andrews, and S. P. Weber, ransmsson capacty of ad hoc networks wth spatal dversty, IEEE rans. Wreless Commun., vol. 7, no. 1, pp , July Engneerng and echnology Publshng K. uang, J. G. Andrews, R. W.. Jr., D. Guo, and R. A. Berry, S spatal nterference cancellaton for mult-antenna moble ad-hoc networks, IEEE rans. Inform. heory, 008. S. Govndasamy, D. W. Blss, and D.. Staeln, Spectral efcency n sngle-hop ad-hoc wreless networks wth nterference usng adaptve antenna arrays, IEEE J. Select. Areas Commun, vol. 5, no. 7, pp , September 007. N. Jndal, J. G. Andrews, and S. P. Weber, Rethnkng mmo for wreless networks: Lnear throughput ncreases wth multple receve antennas, n Proc. IEEE Int. Conf. on Commun, Dresden, Germany, June 009, pp Nhar. Mult-antenna communcaton n ad hoc networks: Achevng mmo gans wth smo transmsson. IEEE rans. Com- mun. [Onlne]. Avalable: O. B. S. Al, C. Cardnal, and F. Gagnon. Performance of optmum combnng n a posson eld of nterferers and raylegh fadng channels. IEEE rans. Commun. [Onlne]. Avalable: M. L. ong, Advances n Multuser Detecton, M. L. ong, Ed., John Wley & Sons, INC., Publcatons, 009. A. Loano and N. Jndal, ransmt dversty vs. spatal multpleng n modern mmo systems, IEEE ransactons on Wreless Communcatons, vol. 9, no. 1, January 010. S. A. Irtaa, A. abb, and Q. ul Islam, Performance comparson of lte transmsson modes n hgh speed channels usng soft sphere decoder, Internatonal Journal of Engneerng & echnology, vol. 1, no. 03, 01, pp Dr. Qamar ul Islam s ead of Department, Electrcal Engneerng, at the Insttute of Space echnology. e has over twenty years of nternatonal eperence n the UK, Mddle East and North Amerca. e s currently leadng wreless and satellte research groups at IS. Dr. Qamar s Chef Edtor of Journal of Space echnology and Project Drector ICUBE-1 Satellte Project. s research nterests are n the area of errestral Wreless Communcaton, Satellte Communcaton and Satellte Engneerng 174