A NOVEL PREAMBLE DESIGN FOR CHANNEL ESTIMATION IN MIMO- OFDM SYSTEMS RESULTING IN ENHANCED THROUGHPUT

Transcription

1 Volume 53, umber 3, 01 ACTA TECHICA APOCESIS Electroncs and Telecommuncatons A OVEL PREAMBLE DESIG FOR CHAEL ESTIMATIO I MIMO- OFDM SYSTEMS RESULTIG I EHACED THROUGHPUT Shakeel Salamat ULLAH atonal Unversty of Scences and Technology (UST), Pakstan shakeelsalamat@ymalcom Abstract: Multple Input Multple Output (MIMO) system combned wth orthogonal frequency dvson multplexng (OFDM) provdes a relable soluton for enhanced data rate next generaton wreless systems as MIMO produces addtonal parallel channels n spatal doman whch allows hgh data rates to be acheved wthout extra bandwdth and transmtted power Whle OFDM provdes hgh bandwdth effcency and robustness to multpath fadng due to orthogonal subcarrers that convert frequency selectve channel nto parallel frequency flat channels Channel estmaton s ntegral part n coherent MIMO-OFDM systems Burst communcaton system utlzes data aded channel estmaton algorthms due to ther fast convergence and accuracy relatve to blnd estmaton algorthms Data aded schemes n MIMO-OFDM employ perodc transmsson of tranng data from transmtter for channel estmaton at the recever In MIMO-OFDM systems hgher number of transmt/receve antennas causes a bgger overhead for channel estmaton Ths paper proposes novel preamble desgn wth low overhead for channel estmaton Smulaton results and analyss of tme based channel estmaton algorthm resultng n spectrally effcent MIMO-OFDM system have also been provded Keywords: Burst Communcaton, preamble desgn, tme based channel estmaton, MIMO-OFDM I ITRODUCTIO ext generaton wreless communcaton system supports hgh data rate applcatons Enhanced throughput and ncreased network capacty are maor requrements of these systems Wth the ncreased number of users, bandwdth allocaton becomes very crtcal Multple antenna technques become a promsng soluton for these requrements Multple Input Multple Output (MIMO) systems ncorporatng Orthogonal Frequency Dvson Multplexng (OFDM) appears to be a match, made n heaven as MIMO-OFDM provdes hgh data rate wth OFDM, ntegrated wth mult-antenna system, reduces the complexty of overall system MIMO provdes parallel channels over the same tme and frequency that can be used to acheve hgh data rates wthout the need of addtonal bandwdth and power MIMO technques can be broadly splt nto space dvson multplexng (SDM) and space tme codng (STC) STC utlzes dversty technques for performance enhancement whereas SDM makes use of multple data streams to acheve hgh data rates OFDM dvdes the frequency selectve channel nto parallel frequency flat channels fghtng multpath fadng and Inter Symbol Interference (ISI) usng Cyclc Prefx (CP) [1] It thus provdes robustness aganst frequency selectve fadng channels [, 3] MIMO-OFDM systems are used n WLA standards and WIMAX and LTE [4, 5, 6] ext generaton of communcaton systems shall be employng MIMO technques to meet the ncreasng data rate requrements and hgh spectral effcency Unlke guded medum, wreless channel causes many unwanted effects to the rado sgnal lke frequency, angle and tme spreadng Often these effects put lmtatons on performance of the systems To reduce these effects channel s estmated and equalzed at the recever Channel estmaton s crtcal n MIMO-OFDM systems, snce n MIMO multple streams are transmtted smultaneously thus error n channel estmates can more crtcally degrade the performance of the overall system Channel estmaton complexty ncreases sgnfcantly wth multple antennas OFDM s therefore generally ntegrated wth MIMO as OFDM converts frequency selectve channel nto ndependent parallel frequency flat channels that makes estmaton of channel at the recever less complex as sngle tap equalzaton s requred The Channel Estmaton (CE) technques generally fall nto two categores Data aded and Blnd channel estmaton Data aded technques for channel estmaton are generally employed n packet based communcaton systems as blnd channel estmaton consumes larger data and tme for convergence In OFDM, channel estmaton can ether be done by nsertng plot tones n all of the subcarrers (Block type CE) or by nsertng plot tones on partcular subcarrers (Comb type CE) nto each OFDM symbol Comb type CE algorthms utlzes nterpolaton technques to estmate channel at remanng subcarrers Block type plot channel estmaton s used n slow fadng channels whereas comb type plot channel estmaton s used to equalze the channel, changng from one OFDM symbol to the other For MIMO-OFDM tranng data used for channel estmaton for partcular transmt/receve antenna should be Manuscrpt receved June 15, 01; revsed July 3, 01 30

2 Volume 53, umber 3, 01 ACTA TECHICA APOCESIS Electroncs and Telecommuncatons orthogonal n tme, frequency or code to other transmt antennas Preamble should be ether tme multplexed, frequency multplexed or code multplexed Smlar technques of channel estmaton for Sngle Input Sngle Output (SISO-OFDM) can be appled to MIMO-OFDM system wth some modfcaton Channel estmaton technques n MIMO systems generally convert the MIMO usng orthogonal tranng sequences nto ndependent SISO The larger the number of transmt antennas, the greater the overhead for channel estmaton as more plot symbols are requred for channel estmaton System model of MIMO-OFDM s provded n secton II Analyss of channel estmaton technques for sngle antenna and multple antenna systems s gven n secton III Preamble desgn wth low overhead for channel estmaton n MIMO-OFDM system s proposed n secton IV and smulaton results and analyss are gven n secton V Concluson s presented n secton VI II SYSTEM MODEL In MIMO-OFDM system, transmsson s made from all transmt antennas smultaneously Let s assume that MIMO- OFDM system operates wth t transmt and r receve antennas Let the sgnal from th transmt antenna, n baseband, s x [ n] 1 k 0 X [ e πk n, g n 1 where g s the number of guard samples or cyclc prefx samples and s the total subcarrers X s the complex data that s modulated onto the subcarrers from th transmt antenna The mult-tap channel between th transmt antenna and th receve antenna s modeled as h ( t, τ ) L l 0 α l ( t) δ ( τ τ ) where L s the total number of complex taps, α l s the l th complex path gan and τ l s the correspondng path delay between th transmt antenna and th receve antenna The receved sgnal y at the th recever antenna s, y [ n] t L 1 1 l 0 h [ l] x [ n l] + n [ n] where h [l] s the l th complex baseband sampled channel tap between th transmt antenna and th receve antenna n s the complex baseband nose at the th recever wth power spectral densty o The guard or cyclc prefx nterval s kept greater than total delay spread of channel The receved sgnal n frequency doman s, Y [ 1 n 0 y [ n] e πn k l,0 k 1 where k s the subcarrer ndex and s the total number of subcarrers For a partcular subcarrer k, the receved sgnal s, (1) () (3) (4) X1[ Y1 [ H11[ H1 t[ X1[ n1[ + (5) Y r[ Hr1[ Hrt[ nr[ X t[ where H [ s the complex baseband channel response at a partcular subcarrer and n [ s the complex nose at a partcular subcarrer k III AALYSIS OF CHAEL ESTIMATIO I MIMO-OFDM SYSTEMS A lot of lterature s avalable on the technques of channel estmaton and equalzaton Least Square (LS), Maxmum Lkelhood Estmaton (MLE), Transform doman and Lnear Mnmum Mean Square Error (LMMSE) estmaton technques are commonly used [7] For orthogonal frequency dvson multplexed systems, tme based and frequency or tone (plot) based channel estmaton technques are studed In tme doman channel estmaton, Channel Impulse Response (CIR) s frst estmated n tme doman and then transformed nto frequency doman for equalzaton at each subcarrer [7] In tone mode, only few subcarrers are used for channel estmaton and then these estmates are nterpolated and extrapolated to get the channel response at the non-plot locatons [8, 9] Wth the ever ncreasng demand for hgh data rates, MIMO systems came nto beng Generally channel estmaton technques for these systems are extended from SISO systems Orthogonal tranng sequence converts MIMO-OFDM channel estmaton to SISO-OFDM, so smlar technques of ML, LS and LMMSE can be used In these channel estmaton technques MIMO-OFDM channel estmaton s done by turnng off the transmt antenna n tme or frequency and sendng plot tones by only one antenna at a tme [10, 11] Data aded channel estmaton technques requre long preambles for hgher number of antennas especally n block type tranng as tranng must be orthogonal n tme for dfferent transmt antennas [1] Comb based technques are proposed for MIMO-OFDM whch requres nterpolaton for null subcarrers as n the case of SISO-OFDM The performance and optmalty of recevers and estmators have remaned under consderable focus ever snce [13,14,15] However, no matchng emphass s seen n lterature on the decrease n overhead of these technques In future systems where MIMO-OFDM wll be playng a key role, overheads for channel trackng wll present a severe bottleneck n achevng hgh throughput for systems nvolvng hgher number of transmt/receve antennas In ths paper ML technque for MIMO-OFDM channel estmaton s presented Preamble s code dvson multplexed, for dfferent transmt antennas, wthn one OFDM symbol duraton ML technque s proposed for channel estmaton In the next secton, preamble desgn s proposed 31

3 Volume 53, umber 3, 01 ACTA TECHICA APOCESIS Electroncs and Telecommuncatons IV PROPOSED PREAMBLE DESIG AD CHAEL ESTIMATIO ALGORITHM 1 ovel Preamble Desgn Proposed preamble desgn for channel estmaton for MIMO-OFDM system s code multplexed Let M be the total number of transmt/receve antenna channels that are to be estmated n one OFDM symbol Channels that can be estmated n one OFDM symbol at maxmum are, M max (6) g Fgure 1 Preamble Desgn for x MIMO-OFDM System where g wll be kept twce of maxmum delay spread In ths case, preamble from th transmt antenna wthout cyclc prefx s, s [ m ( 1) M ] hdm [, ] x [ m ] + (7) where 1 to M, m0 to g -1 and hdm[,] s the th, th element of M hadamard matrx x [m] represents complex baseband data samples n tme doman Data sequence, whch s dfferent for dfferent transmt antennas (as shown n Fg1 and Fg wth dfferent colors), must obey the condton of optmal channel estmaton [16] Complete preamble that ncludes cyclc prefx s formed by copyng last g samples to the begnnng of the OFDM symbol If M s lesser than total number of transmt antennas, more than one OFDM symbol wll be requred n preamble Preamble desgn for x MIMO-OFDM system s shown n Fg1 Preamble desgn for 4x4 MIMO-OFDM shown n Fg n whch t s assumed that g L (8) wth M 4 (9) and 4 g (10) Channel Estmaton usng Proposed Preamble In proposed preamble desgn, data transmsson s made from all transmt antennas smultaneously for the whole OFDM symbol duraton Tme orthogonal tranng sequence causes back off n maxmum transmsson power as tranng sequence transmsson s made from only one transmt antenna at a tme Maxmum lkelhood channel estmaton s proposed for ths novel preamble As the sgnal on a partcular recever antenna wll be the sum of sgnals from all the transmt antennas, sgnal separaton s requred at every recever antenna Sgnal separaton can be made by makng use of the orthogonal coded preamble desgn Each secton of the preamble whch s multpled by a dfferent element of hadamrad marx wll be corrupted by ts prevous secton Thus corrupted porton s cut to make the secton orthogonal at the recever as well Separaton wll then be possble as shown n (11) Fgure Preamble Desgn for 4x4 MIMO-OFDM System h nt [ n] t g p y [ n t p ] hdm[, p] M (11) Where n 0 to M- g / and y [n] s the receved sequence at th antenna after cyclc prefx has been removed h nt [n] s the ntermedate sequence whch s wthout mxng of sgnals and wll be used n channel estmaton between th receve antenna and th transmt antenna CIR can be found usng (1) Wth the ever ncreasng demand for hgh data rates, MIMO systems came nto beng Generally channel estmaton technques for these systems are extended from SISO systems Orthogonal tranng sequence converts MIMO-OFDM channel estmaton to SISO-OFDM, so smlar technques of ML, LS and LMMSE can be used In these channel estmaton technques MIMO-OFDM channel estmaton s done by turnng off the transmt antenna n tme or frequency and sendng plot tones by only one antenna at a tme [10, 11] Data aded channel estmaton technques requre long preambles for hgher number of antennas especally n block type tranng as tranng must be orthogonal n tme for dfferent transmt antennas [1] Comb based technques are proposed for MIMO-OFDM whch requres nterpolaton for null subcarrers as n the case of SISO-OFDM The performance and optmalty of recevers and 3

4 Volume 53, umber 3, 01 ACTA TECHICA APOCESIS Electroncs and Telecommuncatons estmators have remaned under consderable focus ever snce [13,14,15] However, no matchng emphass s seen n lterature on the decrease n overhead of these technques In future systems where MIMO-OFDM wll be playng a key role, overheads for channel trackng wll present a severe bottleneck n achevng hgh throughput for systems nvolvng hgher number of transmt/receve antennas CIR can be found usng (1) h Where S s; 1 [ n] ( S S ) S h (1) H H nt g g x [ ] x[ n] g S,0 n< x[ ] x[ n] M M (13) In (1) t s assumed that (S H S ) s full rank MSE of the channel estmaton s H 1 MSE tr{( S S ) } (14) o TABLE I SIMULATIO PARAMETERS So Parameter Value 1 RF frequency 4 GHz o of Subcarrers 64 3 Subcarrer Spacng 1565 khz 4 Samplng Rate 1Mhz 5 Cyclc Prefx Duraton 16 µs (16 samples) 6 Max Delay Spread 8 µs (8 samples) 7 Max Doppler,f d 8 Carrer Frequency Offset,CFO 9 o of Antennas x,4x4 0,50,100,150,00,50 Hz %,5%,10% of subcarrer spacng antennas, at hgher sgnal to nose rato (SR) preamble overhead can be reduced consderably The smulaton results for dfferent fadng rate or Doppler spread, f d, are shown for x and 4x4 systems n Fg4 and Fg5 respectvely Smulaton results for dfferent carrer frequency offsets and Doppler spread for both x and 4x4 systems are shown n Fg6 and Fg7 The greater the carrer frequency offset (CFO) or the Doppler spread, the greater the MSE CFO has been smulated as percentage of subcarrer spacng of 1565 khz where o s the complex nose power spectral densty and tr{} s the transpose of matrx Thus desgn of S s crtcal and condtons for optmal desgn are gven n [16] Channel can be estmated usng (11), (1) and (13) among M transmt antennas and receve antenna usng one OFDM symbol Channel estmaton s done n tme doman If /M duraton s made exactly equal to g duraton n the preamble, there would be no margn for tmng synchronzaton varance If t s greater than that, more channel estmaton samples wll be avalable whch wll result nto greater margn for tmng synchronzaton V SIMULATIO RESULTS The smulaton parameters are shown n Table1 Mean Square Error (MSE) s calculated for dfferent cases x system and 4x4 MIMO-OFDM system s smulated for dfferent parameters and results are shown Channel taps assumed are 8 The number of samples used n channel estmaton for x system s 3 and n case of 4x4 system, the number of samples used n channel estmaton s 16 Under the parameters gve n Table1 4x4 system s the case, n whch maxmum number of channels (MM max 4) usng one OFDM symbol can be estmated In case of x system, the margn for tmng estmaton varance s hgher as compared to the 4x4 system where more channels are estmated usng the same OFDM symbol 3 samples are used for x system and 16 samples are used for 4x4 system, x system gves us 3dB performance gan as shown n Fg3 However to acheve the same performance for 4x4 system, number of OFDM symbols for channel estmaton must be doubled For hgher number of 33

5 Volume 53, umber 3, 01 ACTA TECHICA APOCESIS Electroncs and Telecommuncatons ncreasng the number of OFDM symbols for estmaton as more samples wll then avalable for channel estmaton REFERECES [1] Zhendao Wang, Gannaks GB, Wreless multcarrer communcaton", IEEE Sgnal Processng Magazne, May 000A [] H Bolcske, D Gesbert, and AJ Paulra, On the capacty of OFDM-based spatal multplexng systems, IEEE Transacton on Communcatons, vol 50, pp 5-34, Feb00 [3] YJ Zhang, and KB Letaef, An effcent resource allocaton scheme for spatal multuser access n MIMO/OFDM systems, IEEE Transacton on Communcatons, vol 53, pp , Jan 005 [4] IEEE 8011a Stand, ISO/IEC :1999/Amd 1:000(E) [5] R van ee, G Awater, M Morkura, H Takanash, M Webster, and K Helford, ew hgh rate wreless LA standards, IEEE Commun Mag, vol 37, pp 8 88, Dec 1999 [6] Further Hgher-Speed Physcal Layer Extenson n the 4 GHz Band, Draft IEEE 8011g Standard [7] Mehmet Kemal Ozdemr, Huseyn Arslan, Channel estmaton for wreless OFDM systems, IEEE Comm Surveys, Vol 9, o, 007 [8] S Coler et al, Channel Estmaton Technques Based on Plot Arrangement n OFDM Systems, IEEE Trans Broadcast, vol 48, pp 3 9, Sept 00 [9] C R Athaudage and A D S Jayalath, Low complexty channel estmaton for wreless OFDM systems, Proc IEEE Int l Symp Personal, Indoor and Moble Rado Commun, vol 1, pp 51 5, Sept 003 [10] W G Jeon, K H Pak, and Y S Cho, An effcent channel estmaton technque for OFDM systems wth transmtter dversty, Proc IEEE Int l Symp Personal, Indoor and Moble Rado Commun, vol, pp , Sept 000 [11] A Dowler, A Doufex, and A x, Performance evaluaton of channel estmaton technques for a moble fourth generaton wde area OFDM system, Proc IEEE Vehc Tech Conf,vol 4, pp , Sept 00 [1] A van Zelst, TCW Schenk, Implementaton of a MIMO OFDM-based wreless LA system, IEEE Transactons on Sgnal Processng, Vol 5, pp , Feb 004 [13] M Hseh and C We, Channel estmaton for OFDM systems based on comb-type plot arrangement n frequency selectve fadng channels, IEEE Transactons on Consumer Electroncs, Feb 1998 [14] L J Cmn, Analyss and smulaton of a dgtal moble channel usng orthogonal frequency dvson multplexng, IEEE Transacton Communcaton, vol33, pp , July 1985 [15] J-J van de Beek, O Edfors, M Sandell, SK Wlson and PO Boresson, On channel estmaton n OFDM systems, n Proc IEEE 45th Vehcular Tech Conf, Chcago, IL, pp M, Jul 1995 [16] Hlang Mnn, Al-Dhahr, Optmal tranng sgnals for MIMO OFDM channel estmaton, IEEE Transactons on Wreless Communcatons,Vol 5, pp , May 006 Fgure 3 MSE Comparson of x and 4x4 MIMO- OFDM system VI COCLUSIO The paper presents the technque for channel estmaton for more than one transmt/receve antenna channels usng one OFDM symbol Proposed method provdes the enhanced throughput capablty wth low overhead, especally wth hgher number of antennas Smulaton results show the performance of the algorthm The performance can be mproved further f M s kept lesser M max Performance mprovement s acheved at the cost of throughput by 34