Co-channel Interference Suppression Techniques for STBC OFDM System over Doubly Selective Channel

Transcription

1 Co-channe Interference Suppression Techniques for STBC OFDM System over Douby Seective Channe Jyoti P. Patra Dept. of Eectronics and Communication Nationa Institute Of Technoogy Rourkea , India E mai : jyotiprasannapatra@gmai.com Poonam Singh Dept. of Eectronics and Communication Nationa Institute Of Technoogy Rourkea , India E mai: psingh@nitrk.ac.in Abstract The Aamouti Space Time Bock Code (STBC) based Orthogona Frequency Division Mutipexing (OFDM) is very attractive in frequency seective channe as the channes remain constant over Aamouti code period (two consecutive OFDM symbo bock). But when the channe is fast fading, this assumption does not hod good and causes co-channe interference (CCI). ence, the simpe Aamouti detection method is not sufficient to recover the origina transmitted signa from the mixed transmitted signas at the receiver side. To cance the effects of CCI, severa detection techniques are addressed namey SIC, DMLD, DZFD, DF and ML. The ML method has the optimum performance but the computationa compexity is very high. Therefore, we propose a suboptima MDF detection which provides comparabe performance to that of ML with very ess computationa compexity. Finay, the performances of the above detection methods are compared in terms of symbo error rate (SER) and computationa compexity for STBC OFDM system over douby seective channe. Key Words STBC, OFDM, CCI, DETECTION I. INTRODUCTION Now-a-days, most of the wireess communication system is focusing on reiabiity and spectra efficiency []-[]. The ink reiabiity is soved by STBC which is proposed by Aamouti [3]. The STBC technique is expanded to arbitrary number of transmit and receive antenna in [4]-[5]. Originay, the STBC technique is appied to fat fading channe. But in practica scenario, the channe is frequency seective as we as time seective rather than fat fading. The frequency seective probem can be soved by appying OFDM technique to the STBC system as OFDM converts the frequency seective fading channe into many narrow parae fat fading channes [6]. owever, in time varying fast fading channe, the Aamouti STBC-OFDM system undergoes two types of interferences namey inter-carrier interference (ICI) and cochanne interference (CCI) [7]-[0]. The ICI occurs due to the oss of orthogonaity among the subcarrier within the OFDM bock. The CCI occurs due to the variation of channe frequency response (CFR) over Aamouti code period. Due to the CCI effect, the two consecutive transmitted OFDM symbo bocks are couped with each other at the receiver side. Both the ICI and CCI cause significant performance degradation in time varying fast fading channe. owever the power of CCI is proved to be 7-8 db greater than the power of ICI regardess of the channe variation [0]. ence, we ony consider the CCI effect on the Aamouti coded OFDM system. In iterature, various detection methods have been addressed for canceing the effects of CCI. In [], a diagonaized maximum ikeihood detector (DMLD) is proposed to remove the CCI. But the DMLD method has more computationa compexity. ence, a simiar performance with ow computationa compexity diagonaized zero forcing detector (DZFD) is proposed in []. A successive interference canceation (SIC) method is proposed in [5]. The decision feedback (DF) method is originay proposed in [3]. Then it is extended for STBC/SFBC OFDM system in [9]. The Maximum Likeihood (ML) detection method are addressed in [9],[]. In this paper, we have anayzed the above detection methods and proposed a suboptima modified DF (MDF) detection method to cance the CCI effect The rest of the paper is organized as foows. In Section II, the system mode for the STBC based OFDM system aong with the channe mode is discussed. The different detection techniques are presented in Section III aong with the computationa compexity. In Section IV, the performance of these detection methods are compared on the basis of SER. Finay, Section V concudes the paper. Input Data Output Data QPSK Moduation QPSK Demoduation on STBC Encoder STBC Decoder Channe Estimation IFFT IFFT FFT Fig. Bock diagram of an STBC-OFDM system mode

2 II. SYSTEM MODEL In this section, an STBC based OFDM system with two transmit antennas and one receive antenna is described aong with the exponentia power deay profie and Jakes sum of sinusoida (SOS) mode as time varying mutipath Rayeigh fading channe. A. STBC Based OFDM System Mode The system mode for STBC-OFDM with two transmit antenna and one receive antenna is shown in the Fig. A random data sequence is generated and moduated according to any specific moduation scheme such as BPSK, QPSK or 6QAM. Then the moduated data are passed through the STBC encoder. The STBC encoder converts the singe input moduated information data into two parae encoded output data according to Aamouti STBC scheme [3]. The Aamouti STBC scheme is described as foows. At the first time instant X and X are transmitted from the transmit antenna and respectivey. In the second time instant X and X are transmitted from antenna and respectivey. ence, the output of the STBC encoder can be written in matrix form and is given as beow X X X X X () k = 0,,, N - X is the k th subcarrier for the transmitting antenna before the IFFT operation and N is the number of subcarriers in a singe OFDM symbo bock. Then, the encoded signa is passed through the seria to parae converter(s/p) and OFDM moduated using Inverse Fast Fourier Transformation (IFFT) bock. Finay, the resuting signas are transmitted from the antennas after insertion of the cycic prefix (CP) which is assumed to be arger than deay spread of the mutipath channe in order to avoid inter symbo interferences (ISI). At the receiver side, at first the CP is removed and then the received signa is demoduated using Fast Fourier Transformation (FFT). The FFT output of the received signa after the remova of CP can be written in the matrix form and is given by Y t,t,t X Y Y ( k) X t T, t T, t T Z Z t t T X Z,t and,t are the CFR for the first and second transmit antennas at the first time instant respectivey. Z t is the circuary compex zero mean additive white Gaussian noise after the FFT operation. () It is to be noted that the channe is taken to be idea it is assumed that the perfect channe knowedge is known at the receiver side. According to Aamouti scheme, the channes are considered to be static over two consecutive OFDM symbo bock, i.e.,, t =, t + T and, t =, t + T. The STBC decoding operation is performed by mutipying on both sides of (), the origina transmitted signa can be recovered after taking the hard decision of the decoded signa and can be written as X ~ Xˆ Q( ) Φ X ~ Φ,t X ~ is the decoded Y,tT signa and Q,t,tT (3) denotes the hard decesion function. ence, the two estimated signas Xˆ are decouped from each other at the receiver side. owever, in the fast fading channe, the channes are not same for two different time sots. ence, is no onger an orthogona matrix and is given by G β β, t, t T, t, t β ( k), t, t T, t T, t T, are the desired diversity gain terms and β, are the CCI terms. By mutipying G with FFT output of the received signa Y, the detected output signa vector can be written as X βx Z X ~ X~ X ~ t (5) β X X Z tt β X and X are the desired signas. βx and β X are the CCI signas which are couped with the desired signa at the receiver side. ence, in order to accuratey recover the origina transmitted signa, these two CCI signas are to be canceed for STBC OFDM in time seective channe. (4)

3 B. Channe Mode In this paper, the wireess channe is assumed to be time varying frequency seective fading (Douby seective) channe. The frequency seective channe is empoyed as exponentia power deay profie (PDP) [4]. The channe is modeed as finite impuse response (FIR) with tota L+ non-zero path with zero mean and average power. The channe can be expressed as h N(0, ) jn(0, ) (6) N(0, ) is the zero mean with variance. The power of mutipath component decreases exponentiay. The first path of the mode is chosen to be λ 0 λ e L T s τ rms (7) λ The T s and τ rms are the samping period and root mean squared deay of the channe respectivey. The energy of the th path can be written as 0 λ. (8) Furthermore, each mutipath is modeed as uncorreated Rayeigh fading channe with Jakes sum-of-sinusoida (SOS) mode [5]. III. DETECTION METODS FOR CCI CANCELLATION In this section, we describe different detection methods namey the SIC, DZFD, DF, ML and the proposed MDF methods in order to cance the effects of CCI in the STBC based OFDM system over douby seective channe. A. SIC Detection Method The successive interference canceation method is proposed in [0]. Due to the time varying nature of channe, the two diversity gain terms are not same but the gains of the two CCI terms are equa as given in (4). ence the SIC detection method is based on this gain difference property to recover the origina transmitted signa and is iustrated beow. X ~ X βx Zt (9). X ~ β X X Z tt (0) 3. if Xˆ Q(X ~ ) Xˆ Q((X ~ β Xˆ )/ ) () 4. ese Xˆ Q(X ~ ) Xˆ Q((X ~ βxˆ )/ ) () B. DZFD Detection Method In the time varying fast fading channe, mutipying with does not give a orthogona matrix as expained in (4). In order to get an orthogona matrix, a transform matrix Ω is mutipied with the matrix [] and is given beow Ω diag(, ) (3) and are the compex number. After simpification of (3), the Ω can be written in the matrix form and is given beow,t, t T Ω (4), t T, t The and have the same vaue and is given by (5), t T, t, t, t T Substituting Ω in the pace of, the decoded signa X ~ can be written as X ~ ΩY diag(, )X ΩZ (6) The estimated origina transmitted signa can be obtained by dividing the vaue of on both side of (6) and then taking the hard decision X ~ Xˆ Q( ) X Z (7) X ~ Xˆ Q( ) X Z (8) The DZFD detection method is simpe as it simpy forces the two CCI terms to zero. C. DF Detection Method The decision feedback (DF) detector was proposed for STBC-OFDM system in [9]. The working principe of DF detection method is described as foows. The DF first estimated a signa and then uses this estimated signa to hep make decision about the other estimated signa by assuming that the first estimated signa is correct. The agorithm of DF detection method is given beow.. Estimate the first signa Xˆ using DZFD detection method by using the (7).. Cance the contribution of Xˆ by subtracting it from the mixed detected output signa in (9). Then take the hard decision to obtain the estimated signa. The estimated signa Xˆ can be written as Q((X ~ Xˆ β Xˆ )/ ) (9)

4 D. Proposed Modified DF (MDF) Detection Method In this paper, we have proposed a modified DF (MDF) detection to cance the CCI effect. This method is the combination of SIC and DZFD method. The MDF method chooses the estimated signa coming from DZFD method based on gain diversity property and then cance their contribution according to SIC method with some modification. The agorithm for MDF detection method is given beow. if Xˆ Q((Xˆ / ) (0) Q((X ~ Xˆ β X ~ )/ ) () Xˆ Q((X ~ βxˆ )/ ) (). ese Xˆ Q((Xˆ / ) (3) Xˆ Q((X ~ βxˆ )/ ) (4) Xˆ Q((X ~ β Xˆ )/ ) (5) E. ML Detection Method The ML detection estimates a possibe combination of transmitted signa and chose the most probabe one. The ML detection can be written mathematica as Xˆ ( k) arg min X X C CM, M X Y ( k) ( k) X denotes the consteation points. ML detection requires CM (6) times matrices cacuation to sove (3) in order to estimate the origina transmitted signa in the receiver side. The ML detection performs the best among a the detection method but with the highest computationa compexity ( CM matrices cacuation) and hence is not preferabe from the hardware impementation perspective. Computationa Compexity Tabe.Computationa Compexity Aamouti SIC DZFD DF MDF ML (C M ) QPSK QAM F. Computationa Compexity To cacuate the computationa compexity, we consider ony compex mutipication and divisions per subcarrier. Aamouti detection method invoves 4 compex mutipications for X ~ = Y. In addition to that and invoves mutipication. X~ / and X~ / does not require any compexity as both and are scaar quantity. For SIC detection technique 4 compex mutipication are required for X ~ and,, β requires each two compex mutipication. Xˆ β and and β Xˆ needs each one mutipication each. Division by does not require any computationa compexity. Therefore, tota computationa compexity invoved in SIC method is. In DZFD, mutipying Ω with Y requires 4 compex mutipication and needs compex mutipication. It is noted that one compex division require twice the number of compexity to compex mutipication. ence, DZFD invoves tota 0 numbers of compex mutipications. DF detection technique needs 0 compex mutipication to estimate Xˆ using DZFD method, 4 compex mutipication for Y and 6 compex mutipication for β,,. Aso one compex mutipication is required for β Xˆ.The MDF invoves same number of computationa compexity to that DF with additiona 3 compex mutipication i.e. number of β Xˆ and one β Xˆ. Tota number of computationa compexity for MDF method is 4. The ML detection method invoves 4(C M ) compex mutipications to sove (6). The computationa compexity of various detection methods are summarized in Tabe. IV. SIMULATION RESULTS The performance comparison of different detection methods incuding Aamouti, SIC, DZFD, DF, MDF and ML methods for STBC based OFDM system are carried out on the basis of SER over various mobie speeds. The simuation parameters used for simuation are isted on the Tabe. Tabe.Simuation Parameter Parameter Vaue FFT Size 8 Number of Subcarrier 8 Number of CP 3 Data Moduation QPSK Carrier Frequency 5Gz Channe BW Mz Subcarrier Spacing 7.85Kz Channe Mode Exponentia decaying PDP Number of Mutipath 4 RMS Deay spread us Mobie speed 50/00/00 Km/h Normaized Dopper 0.03/0.06/0. frequency (F d NTs) The SER performances for SISO OFDM and STBC OFDM with different CCI canceation techniques e.g. Aamouti, SIC, DZFD, DF, ML and MDF detectors for mobie speed 50Km/h (normaized Dopper frequency F d NT s = 0.03) is shown in the Fig.. The simuation resuts show that STBC OFDM with Aamouti method performs better than SISO OFDM but this method does not provide sufficient performance as it suffers

5 Symbo Error Rate Symbo Error Rate Symbo Error Rate severe performance degradation due to both CCI and ICI. The performance of SIC method is better than Aamouti detection method but is not accurate enough due to error propagation probem. The DZFD detector simpy forces the CCI signa to zero and gives better resuts than SIC method. But it suffers from noise enhancement. The DF gives better resut than Aamouti, SIC and DZFD methods. The MDF outperforms a the detection methods as it takes the advantages of both DZFD and SIC method but its performance is esser than ML method. The computationa compexity of various detection methods are given in Tabe. From the tabe, it is seen that ML method has the highest as Aamouti method has the east computationa compexity foowed by DZFD, DF and MDF. We aso see that MDF method achieves resuts comparabe to the ML method with esser computationa compexity SISO OFDM Aamouti SIC DZFD DF MDF ML SNR in db Fig. SER Vs SNR for different detection methods for normaized dopper frequency of 0.03 SISO OFDM Aamouti SIC DZFD DF MDF ML SNR in db Fig.3 SER Vs SNR for different detection methods for normaized dopper frequency of 0.06 Fig.3 and Fig.4 show the SER performance of various detection methods for normaized Dopper frequency of 0.06 and 0. respectivey. The simuation resuts show that as the normaized Dopper spread increases the Aamouti STBC OFDM becomes worse than SISO OFDM at high end of the SNR. From the Fig.-Fig.4, we have seen that, the performance of a the detection methods decreases drasticay with increase in vehice speed. The performance of different detection methods in the descending order are given as foows: ML, MDF, DF, DZFD, SIC and Aamouti SISO OFDM Aamouti SIC DZFD DF MDF ML SNR in db Fig.4. SER Vs SNR for different detection methods for normaized dopper frequency 0. V. CONLCUSIONS In this paper, we have anayzed various detection methods to suppress the effects of CCI for STBC OFDM system over douby seective channe and proposed a MDF detection technique. The various detection techniques which incude Aamouti, SIC, DZFD, DF, ML and the proposed MDF method are compared in terms of SER and computationa compexity. From the simuation resuts, it is seen that the proposed MDF method achieves comparabe performance to the ML method with a esser computationa compexity. REFERENCES [] G. Stuber, Broadband MIMO-OFDM Wireess Communications, IEEE Transaction on Signa Processing, vo.9, pp. 7-94, 004. [] A. F.Naguib, N. Seshadri and A. R Caderbank, Increasing data rate over wireess channes: Space time coding and signa processing for high data rate wireess communications, IEEE Communications Magazine, vo.7, pp.76 9, 000.

6 [3] S. M Aamouti, A simpe transmitter diversity scheme for wireess communications, IEEE Journa of Seected Areas in Communications, vo.6, pp , 998. [4] V. Tarokh,. Jafarkhani and A. R. Caderbank, Space time bock codes from orthogona designs, IEEE Transaction on Information Theory, vo. 45, pp , 999. [5]. Jafarkhani, A quasi-orthogona space time bock code, IEEE Transactions on Communications, vo. 49, pp. 4, 00. [6] K. F. Lee and D. B. Wiiams, A space-time coded transmitter diversity technique for frequency seective fading channes, IEEE Sensor Array and Mutichanne Signa Processing Workshop, Cambridge, MA, pp. 49-5,000. [7] K. I. Lee,. I. Yoo, J. Kim and Y. S. Cho, Computationa efficient signa detection for STBC-OFDM systems in fast-fading channes, IEEE 8th Internationa Symposium on Persona, Indoor and Mobie Radio Communications, 007. [8] C. Y. Tso, J. M. Wu and P. A. Ting, Iterative interference canceation for STBC-OFDM systems in fast fading channes, IEEE Goba Teecommunications Conference, pp. -5, 009. [9] D. B. Lin, P.. Chiang and. J. Li, Performance anaysis of two branch transmit diversity bock-coded OFDM systems in time-varying mutipath Rayeigh-fading channes, IEEE Transaction on Vehicuar Technoogy, vo. 54, pp , 005. [0] J. W. Wee, J. W. Seo, K. T. Lee, Y. S. Lee and W. G. Jeon, Successive Interference Canceation for STBC-OFDM systems in a Fast Fading Channe, IEEE Vehicuar Technoogy Conference, pp ,005. []. Kanemaru and T.Ohtsuki, Interference canceation with diagonaized maximum ikeihood decoder for spacetime/space-frequency bock coded OFDM, IEEE Vehicuar Technoogy Conference, pp ,004. [] C.M. Li, I.T. Tang and G.W. Li Performance Comparison of the STBC OFDM Decoders in a Fast Fading Channe, Journa of Marine Science and Technoogy, vo.0, vo. 5, pp , 0. [3] Y. Li Viemon, and J. R. Barry, Performance of transmit diversity over time-varying Rayeigh-fading channes, IEEE Goba Communications Conference, pp , 00. [4] G. L. Stüber, Principes of Mobie Communication, London: Kuwer, 00. [5] W. C. Jakes, Microwave Mobie Channes, New York: Wiey, 974.