Australian Journal of Basic and Applied Sciences. Performance Analysis of Pilot Based Channel Estimation Techniques In MB OFDM Systems

Transcription

1 AENSI Journals Australian Journal of Basic and Alied Sciences ISSN: Journal home age: Performance Analysis of Pilot Based Channel Estimation Techniques In MB OFDM Systems Dr. C. Venkatesh Associate Professor, Deartment of Electronics and Communication Engineering, Sri Eshwar College of Engineering, Coimbatore , Tamilnadu, India. A R T I C L E I N F O Article history: Received 25 June 2014 Received in revised form 8 July 2014 Acceted 10 August May 2014 Available online 30 August 2014 Keywords: UWB, OFDM, Least Square, Wireless Channel, Least Minimum Mean Square A B S T R A C T Background: Performance of the ilot based channel estimation techniques has been analyzed for Multiband OFDM systems. Objective: To roose a new ost rocessing aroach to imrove the erformance of channel estimator. Results: : The simulation results that the MB OFDM system rovides considerable results for vehicular and edestrian channel models. Conclusion: In this aer, some of the key issues for design of multicarrier UWB communications have been analyzed. The erformance of MB OFDM system has been analyzed using LS and MMSE based channel estimation algorithm by considering EPA, EVA and ETU channel models with different interolation techniques. It has been observed from the simulation results that the MB OFDM system rovides considerable results for vehicular and edestrian channel models. This work can be extended by considering various caacity enhancement techniques in the OFDM system AENSI Publisher All rights reserved. To Cite This Article: Dr. C. Venkatesh., Performance Analysis of Pilot Based Channel Estimation Techniques In MB OFDM Systems. Aust. J. Basic & Al. Sci., 8(13): , 2014 INTRODUCTION UWB communications has received great interest from both the research community and industry in recent years. The restriction of transmitter ower has motivated the research in UWB communication for short range alications articularly in Sensor networks (SN) and Wireless Personal Area Networks (WPANs). Several GHz of bandwidth has been authorized for license free communication by the Federal Communication Commission (FCC) in United States (US). FCC has mandated that the UWB radio transmission is lies between 3.1 to 10.6 GHz with a minimum instantaneous bandwidth of 500MHz (or) fractional bandwidth in excess of 20% of center frequency. The rincile of OFDM is to slit a high rate data stream into a number of lower rate streams which are transmitted simultaneously over a number of subcarriers. The interference between two symbols is avoided in OFDM transmission by the use of cyclic refix (CP). It has been recently alied in Wireless Local Area Networks (WLANs) for high data rate communications. UWB OFDM communication was roosed for hysical layer in the IEEE a standard which covers wideband communication in Wireless Personal Area Networks (WPANs)( Saleh, A. and R. Valenzuela, 1987). In this aer, there are three channel models considered which suits for edestrian, vehicular, and tyical urban roagation scenarios. The channel models are Extended Pedestrian-A (EPA), Extended Vehicular-A (EVA) and Extended Tyical Urban (ETU). The detailed exlanation for channel models is given in Chater 2. There are two main roblems are considered while designing channel estimators for wireless OFDM systems. The first roblem is, arrangement of ilot information over one OFDM symbol. The ilot signal means the reference signal used by the transmitter and it is known at the receiver. The second roblem is the design of an estimator with low comlexity and good channel tracking ability. Based on the channel conditions, the channel estimation can be erformed by either inserting ilot tones into all of the subcarriers of OFDM symbols with a secific eriod or inserting ilot tones into each OFDM symbol. The first method of channel estimation is called block tye ilot channel estimation. This method has been develoed for slow fading channels. The estimation of the channel for this block-tye ilot arrangement can be based on Least Square (LS) or Minimum Mean-Square Error (MMSE). The MMSE estimate gives gain in Signal-to-Noise Ratio (SNR) for the same mean square error of channel estimation over LS estimate. The combtye ilot channel estimation has been introduced to equalize the effect of fast time varying channel. This tye Corresonding Author: Dr. C. Venkatesh, Associate Professor, Deartment of Electronics and Communication Engineering,, Sri Eshwar College of Engineering, Coimbatore , Tamilnadu, India. E-mil id:cvenkateshmail@yahoo.com, Tel :

2 722 Dr. C. Venkatesh, 2014 of channel has the variation even within OFDM block. The comb-tye ilot channel estimation consists of algorithms to estimate the channel at ilot frequencies and to interolate the channel. The key arameters for UWB channel environments are described by Molisch (2003). Liano et al.(2009) has reorted the arameters of UWB channel model based on frequency domain aroach with lognormal statistics. It is very useful for UWB roagation, leading to the derivation of more accurate channel models in both system design and erformance otimization. The analysis of channel estimation algorithms is resented in Saqib Saleem and Qamar-Ul-Islam (2011). In this aer, the erformance of LS and LMMSE channel estimation techniques is analyzed for different channel imulse resonse samles. Neetu Sood et al.(2010) has reorted on channel estimation of OFDM for Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK) modulation techniques. Keeing the above facts, LS and LMMSE based channel estimation methods are considered in this aer. The estimated channel resonses are filtered using time domain low ass filter in which the channel delay sread is assumed as known arameter. The filtered channel coefficients are converted into frequency domain and the interolation is erformed. The interolation of the channel estimation can deend on linear interolation, second order interolation, low-ass interolation, Sline cubic interolation and time domain interolation. The erformance of MB OFDM system with LS and MMSE based ilot channel estimation techniques with various interolation techniques has also been analyzed in this aer and results are given in results and discussions. System Model and Channel Model: MB OFDM System Model: The generalized block diagram of Multiband OFDM (MB OFDM) system is shown in Fig.1.The binary information is the first groued and maed according to the modulation in the signal maer. The maed signals are converted into arallel which is more efficient for high data rate communication. Pilot ositions are selected based on the channel conditions and accordingly ilot bits are inserted with the data bits. There are two kinds of channel estimation techniques that can be emloyed in OFDM based systems namely Block tye and Comb tye. In Block-tye ilot based channel estimation, OFDM channel estimation symbols are transmitted eriodically, in which all sub-carriers are used as ilots. If the channel is constant during the block, there will be no channel estimation error since the ilots are sent at all carriers. In Comb-tye ilot based estimation, ilot bits are inserted uniformly across the entire bandwidth (Mohammed Safiqul Islam et al., 2011; Carlos Augusto Rocha et al., 2007). Fig. 1: Block diagram for OFDM system The ilot signals are uniformly inserted into according to the following equation X ( k) X ( ml l) where x( m), l 0 Data, l 1,...( L1) (1)

3 723 Dr. C. Venkatesh, 2014 L is the number of carriers in which the ilot bits are inserted uniformly. x ( m) is the th m ilot carrier th value and X( k ) is the frequency domain value at the k subcarrier of the OFDM symbol. After inserting ilots either to all sub-carriers with a secific eriod or uniformly between the information data sequence, IFFT block is used to transform the data sequence of length N into time domain sequence. The frequency domain value xk ( ) is transferred into time domain and it canbe written as, x( n) IFFT X( k) n0,1,2... N 1 N 1 j(2 kn/ N ) X k e (2) x( n) ( ) k 0 where N is the FFT/IFFT length and xn ( ) is the time domain value of the samle n. After IFFT block, Cyclic Prefix (CP) is inserted in order to avoid inter symbol interference. Normally the length of the cyclic refix deends on the channel conditions. The length of cyclic refix should be always greater than the delay sread of the channel. The insertion of CP will avoid Inter-Carrier Interference (ICI) by maintaining the orthogonality across all the subcarriers. The resultant OFDM symbol can be reresented as, x( N n), n NCP, NCP 1,...,1 xf ( n) x( n), n 0,1,..., N 1 (3) where N CP is the number of cyclic refix samles inserted in the OFDM symbol. The CP inserted OFDM symbols are assed over the wireless channel. The channel imulse resonse h(n) is convolved with transmitted signal xn ( ) and it is added with noise signal wn ( ) The received signal can be reresented as, y( n) h( n)* x( n) w( n) (4) The received signal is converted serial into arallel stream and guard interval is removed and it is converted in to frequency domain signal Yk ( ), it is reresented as Y( k) FFT y( n) k 0,1,2,..., N1 1 Y ( k) ( ) N N 1 j(2 kn/ N ) y n e (5) n0 Assume there is no ISI due to the channel, the resultant signal Yk ( ) can be related as Y( k) X ( k) H( k) I( k) W( k) k 0,1,2,..., N 1 (6) where Hk ( ) is the Frequency Domain (FD) reresentation of CIR and Ik ( ) is the interference art. The ilot symbols are extracted from the received symbol Yk ( ) and channel estimation rocess is erformed. H ( k) is the estimated channel resonses which can be defined as, H ( k) e Y ( k) (7) X ( k) where Y ( k ) and X ( k ) are the ilot signal values at the transmitter and the receiver. The frequency domain channel resonses for the data subcarriers are derived from H ( k ) using interolation oeration. The transmitted data symbols are estimated using the interolated channel resonses and it is finally demaed at the receiver (Riazul Islam, S.M. and Kyung Su Kwak, 2010; Mehmet Kemal Ozdemir and Hueyin Arslan, 2007). The received symbols are equalized using interolated channel resonses and given to the modulation demaer. e e

4 724 Dr. C. Venkatesh, 2014 Channel Model: In general, an arbitrary realization of the channel can be reresented with a finite imulse resonse filter as, N 1 h( t) r ( t r) (8) where l0 is the ath gain, Δ is the multiath resolution and N is the maximum delay sread. When () r assed through the multiath channel, the received signal can be reresented as N 1 r( t) x( t)* h( t) rx( t r) (9) l0 xt is The evaluations of wireless standards demand the channel models with increased bandwidth to reflect the characteristics of the radio channel. The resolution of the receiver deends on the channel bandwidth (Zhong Wang et al., 2010; Raffaello Tesi et al., 2007). The latest wireless standards such as LTE (Long Term Evolution) channel models were based on a synthesis of existing models such as the International Telecommunication Union (ITU) channel models (Yun Liu, Qicong Peng et al., 2009; Melisa Barrera et al., 2009). Secifically the six ITU models covering maximum delay sread from 35 ns to 4000 ns were considered. In this way, extended wideband models with low, medium, and large delay sread values are identified in this aer. The low delay sread gives an Extended Pedestrian A (EPA) model which is emloyed in an urban environment with fairly small cell sizes. This channel model can be considered even u to 2 km in suburban environments with low delay sread (Wen Zhou and Wong-Hing Lam, 2010). The medium and large delay sreads give an Extended Vehicular A (EVA) model and Extended TU (ETU) model resectively. The ETU model has a large maximum excess delay of 5000 ns which in fact is not very tyical in urban environments. Instead it alies to some extreme urban, suburban, and rural cases which occur seldom but which are imortant in evaluating LTE erformance in the most challenging environments (Hussein Hijazi and Laurent Ros, 2009; Xenofon G. Doukooulos and George V. Moustakides, 2004). The r.m.s. delay sread values for the three extended models is reresented in Table 1. It was also decided that the extended channel models are alied with low, medium and high Doler shifts, namely 5 Hz, 70 Hz and 300 Hz, which at a 2.5 GHz carrier frequency corresond roughly to mobile velocities of 2, 30 and 130 km/h resectively. Combinations which are likely to be used are EPA 5 Hz, EVA 5 Hz, EVA 70 Hz and ETU 70 Hz. Table.1: Proagation conditions for multiath fading in the ITU channel models. Ta Number ITU edestrian A ITU edestrian A ITU Vehicular A Relative delay (ns) Relative mean ower (db) Relative delay (ns) Relative mean ower (db) Relative delay (ns) Relative mean ower (db) Channel Estimation in OFDM System: Two basic one dimensional channel estimations in OFDM systems are illustrated in Fig.2. In block tye channel estimation is develoed under the assumtion of slow fading channel. It is erformed by inserting ilot tones into all subcarriers of OFDM symbols within a secific eriod. In comb-tye ilot channel estimation which is introduced to satisfy the need for equalizing when the channel changes even from one OFDM block to the subsequent one. It is erformed by inserting ilot tones into certain subcarriers of each OFDM symbol where the interolation is needed to estimate the conditions of data subcarriers. In this aer, block tye channel estimation technique is considered. The LS and MMSE based channel estimation methods are considered. Least Square Channel Estimation: The LS estimation method is the simlest and comutationally least comlex method of obtaining the channel estimates. Since the transmit as well as the received data are known at the reference signal locations, the LS channel estimates at those locations are given by, H ( k) e Y ( k) (10) X ( k)

5 725 Dr. C. Venkatesh, 2014 Fig. 2: Different tyes of ilot arrangement. V( k) He( k) H ( k) X ( k) (11) V ( k) The variance of He( k) deends on the variance of. Both the variances are same in case of BPSK X ( k ) modulation. The LS method is comutationally chea but it has higher estimation Mean Squared Error (MSE). MMSE Channel Estimation: The standard alternative to the Least Squares (LS) method is the Minimum Mean Squared Error (MMSE) method for obtaining the channel estimates. The MMSE method also considers the channel statistics as well as the noise statistics for obtaining the estimates which results in an imroved channel estimate. The estimation of channel frequency resonse as er wiener filter equation is, H H H HH g g gg (12) R E HH E F F FR F where g is the vector consists of time domain channel coefficients. H is the vector consists of frequency domain channel coefficients. H FFT ( g) H H H H 2 H H H gy g gg R E gy E g XF N R F X (13) R E YY XFR F X I (14) YY gg N N where R gg R, HH and YY R is the auto covariance matrix of gh, and Y resectively. covariance matrix between g and Y resectively. Rgy 2 N is the noise variance of white gaussian noise. Assume the channel vector g and the noise N are uncorrelated and receiver in advance, the MMSE estimator of g is given by, R gg and is the cross 2 N are known at the gˆ R R y 1 MMSE gy yy HH (15) The mean square value of linear estimator is given by, Hˆ Fg F[( F X ) R XF] Y H H MMSE MMSE gg N H H FR [( ) ] ˆ gg F X XF N Rgg F HLS [ ( H R ) ] ˆ HH RHH N XX HLS (16)

6 BER 726 Dr. C. Venkatesh, 2014 The MMSE estimator gives much better erformance than LS estimators articularly under low SNR scenarios. A major drawback of the MMSE estimator is its high comutational comlexity in case of matrix inversions are needed each time when the data changes. In time domain windowing aroach, the estimated channel resonses H are converted into time domain and multily with a window. The length of the window LS is equal to the delay sread of the channel and it is also exressed as, H H w (17) modified LS where w is the time domain filter and w = [(ones(1,length(delay sread)) zeros(no. of ilots-length(delay sread)]. The modified channel estimates are given to inut to the MMSE module. RESULTS AND DISCUSSIONS Table 2 resents the system arameters are considered for simulations. synchronization has been assumed for estimating the channel erformance. In this aer the erfect Table 2: Selected simulation arameters. Parameter Value Modulation BPSK/QPSK/QAM FFT Size 1024 Cyclic Prefix (CP) Length 1/16 Signal to Noise Ratio (SNR) 0 to 50 db Pilot Density 1:1/1:3/1:7 Velocity (Km/Hr) 700 Carrier Frequency 1 GHz Interolation Linear/chi/sline Channel Estimation LS/LMMSE The erformance analysis of MB OFDM systems over Additive White Gaussian Noise channel for different hase shift keying techniques is shown in Fig.3. It shows that, the erformance of the 64 PSK is 5 to 10 db better comared with 32 PSK BPSK QPSK 8PSK 16PSK 32PSK 64PSK SNR Fig. 3: BER Performance analysis for different PSK modulation techniques. The erformance of BER for EPA,EVA and ETU channel models is shown in figure 4. It can be seen that, for EPA channel model the BER is low comared with EVA and ETU channel models. This is because there is a variation in the channel delay sread.

7 BER BER 727 Dr. C. Venkatesh, EPA EVA ETU SNR Fig. 4: BER erformance of EPA, EVA and ETU channel models channel. The Bit Error Rate(BER) erformance has been analyzed for different ilot density atterns is shown in Fig. 5. The ilot density atterns which are considered are 1:1,1:3 and 1:7. 1:1 means, for each data subcarriers there will be one ilot. In fig. 5, the erformance of the system imroves where there is high number of ilot for the cost of less throughut efficiency PD-1:1 PD-1:3 PD-1: SNR Fig. 5: BER erformance of different ilot density atterns.

8 BER 728 Dr. C. Venkatesh, Linear Pchi Sline SNR Fig. 6: BER erformance of various interolation techniques channels. The erfomance of various interolation techniques based on bit error rate is shown in figure 6. The figure ortrays that chi interolation gives the better erformance comared with linear and sline interolation techniques. Fig. 7: MSE analysis of LS and MMSE algorithms channels. The erfomance of LS and MMSE channel estimation techniques based on mean square error is shown in figure 7. The figure ortrays that, based on the mean square error MMSE gives the better erformance comared with LS estimation technique.

9 729 Dr. C. Venkatesh, 2014 Fig. 8: Performance comarison of channel estimation schemes between LS and roosed methods. Figure 8 shows comarison of the redicted channel co-efficients using LS method and filtered method in terms the normalized mean square error. It ortrays that NMSE of the filtered one gives 5dB better erformance than the LS method. Conclusion: In this aer, some of the key issues for design of multicarrier UWB communications have been analyzed. The erformance of MB OFDM system has been analyzed using LS and MMSE based channel estimation algorithm by considering EPA, EVA and ETU channel models with different interolation techniques. It has been observed from the simulation results that the MB OFDM system rovides considerable results for vehicular and edestrian channel models. This work can be extended by considering various caacity enhancement techniques in the OFDM system. REFERENCES Carlos Augusto Rocha et al., Performance Analysis of Channel Estimation Schemes for OFDM Systems, Proc. IWT-07, Gonzalo Liano et al., The UWB-OFDM Channel Analysis in Frequency, IEEE Latina America Transactions, 7(1): Hussein Hijazi and Laurent Ros, Joint data QR-Detection and Kalman Estimation for OFDM Time Varying Rayliegh Channel Comlex Gains, IEEE Transactions on Communications, 58(1): Mehmet Kemal Ozdemir and Hueyin Arslan, Channel Estimation for Wireless OFDM Systems, IEEE Communications, 9(2): Melisa Barrera et al., A Novel SNR Estimation Algorithm for MB OFDM Ultra Wide Band communications, IEEE transaction on Communication, 50(3): Mohammed Safiqul Islam et al., Performance Analysis of Different Modulation Techniques in Rayleigh Fading Channel, IJFPS, 1(1): Molisch, A.F., Channel Models for Ultra Wideband Personal Area Networks, IEEE Wireless Communications, Neetu Sood et al., On Channel Estimation of OFDM-BPSK over Nakagami-m Fading Channels, SPIJ, 4(4): Raffaello Tesi et al., On the Performance Comarison of Different UWB Data Modulation Schemes in AWGN Channel in the Presence of Jamming IEEE Trans. Communications, 2: Riazul Islam, S.M. and Kyung Su Kwak, Winner-Hof Interolation Aided Kalman Filter- Based Channel Estimation for MB-OFDM UWB Systems in Time Varying Disersive Fading Channel, ICTACT. Saleh, A. and R. Valenzuela, A Statistical model for indoor wireless multiath roagation, IEEE JSAC, Vol SAC-5(2): Saqib Saleem and Qamar-Ul-Islam, Performance and Comlexity Comarison of Channel Estimation Algorithms for OFDM Systems, IJECS-IJENS, 11(2): 6-12.

10 730 Dr. C. Venkatesh, 2014 Wen Zhou and Wong-Hing Lam, Channel Estimation and Data Detection over Fast Fading and Disersive Channels, IEEE Transactions on Vehicular Technnology, 59(3): Xenofon G. Doukooulos and George V. Moustakides, Blind Adative Channel Estimation in OFDM Systems IEEE ICC 2004, , Yun Liu, Qicong Peng et al., Blind Channel of Subcarrier Number in Multiband OFDM Ultra Wide Band Communications System IEEE Proceedings of IC-NIDC, Zhong Wang et al., A Low-Comlexity and Efficient Channel Estimator for Multiband OFDM- UWB Sytems, IEEE Transactions on Vehicular Technnology, 59(3):