Integer Frequency Offset Algorithm for Digital Radio Mondiale System

Transcription

1 Journal of Communications Vol 8, o 9, September 013 Integer Frequency Offset Algorithm for Digital Radio Mondiale System Cheng Yan and Ming Yan Communication University of China, Beijing, China yancheng66@aliyuncom; yanming668@gmailcom; Abstract This paper proposes one frequency offset estimation algorithm for orthogonal frequency division multiplexing (OFDM) based digital radio mondiale (DRM) system The algorithm exploits the correlation of frequency pilots to construct a novel angle vector As the power factor of the pilots in the correlated calculation is inevitably disturbed by the multipath and noise, the angle vector only utilizes the phase factor of the pilots The frequency offset can be obtained by detecting the shift of the vector norm of the pilots, because the vector norm is maximum at the position of the pilots The performance of the proposed algorithm is compared with that of conventional algorithms The simulation results show that the proposed algorithm can effectively combat multipath and noise with wider range and higher accuracy of the frequency offset estimation Index Terms Digital radio mondiale, orthogonal frequency division multiplexing, frequency offset I ITRODUCTIO The DRM system utilizes OFDM modulation [1] OFDM as an orthogonal multi-carrier modulation system is much more sensitive to carrier frequency offset than the single-carrier system As the carrier frequency offset can destroy the orthogonality of sub-carriers and bring the inter-carrier interference (ICI), the frequency synchronization is particularly important to OFDM system The carrier frequency offset (CFO) estimation is usually divided into two parts: frequency offset (IFO) []-[4] and al frequency offset (FFO) [5]- [9] ots of research has been done for OFDM frequency synchronization algorithms Basically, they are summarized into two categories: the training sequence assisted and non-data assisted The non-data assisted algorithm, for example, M (maximum likelihood) based on CP (cyclic prefix) With the help of the correlation between the CP and the back portion of the OFDM symbol, the algorithm can complete the symbol synchronization estimation [10] As the frequency offset estimation range 05 which is normalized to the sample interval, the algorithm can t do frequency offset estimation [11] The method based on the training sequence is the joint estimation of time and frequency To get accurate Manuscript received July 1, 013; revised September 1, 013 Corresponding author yancheng66@aliyuncom doi:10170/jcm frequency offset, the estimation of time-delay and the al frequency offset are necessary S&C [1] is proposed by T M Schmidl and D C Cox, in which the presence of peak platforms leads to the vague timing synchronization Minn [13] re-designs the first two OFDM symbols, and the algorithm eliminates the flat effect of the timing metric function However, the method has a large root mean square error in multipath channel, and in a quarter of the cyclic prefix to the number of subcarriers, the performance is not satisfactory Park [14] proposes one synchronization algorithm which can produce a sharp peak However it would be accompanied by side peaks Conventional synchronization algorithm based on the correlated property of the time-domain pseudo-noise (P) sequence may not work well in multipath channels An algorithm is proposed in [15] where the P sequence in the frame head is considered as a CP of the OFDM training symbol The frequency synchronization algorithm based on the training sequence develops the synchronization performance by increasing the system overhead with the low spectrum utilization and limited estimated range This paper will focus on the frequency offset estimation algorithm for DRM receiver [16] Basically, they are summarized into two categories: the time pilot reference and frequency pilot reference The problem of frequency offset estimation has been widely explored, but there is still room for a better estimator which has a wider range and higher accuracy The section II mainly introduces the DRM system model The section III describes the conventional frequency synchronization algorithm for DRM system The section IV will introduce the proposed algorithm Finally, the proposed algorithm is verified by simulation II THE DRM SYSTEM MODE The DRM system utilizes the OFDM modulation At the transmitter, an OFDM symbol x( n), n[ g, 1], is generated by performing an -point inverse fast fourier transform (IFFT) on the information symbol X(k) for k [ kmin, kmax ] and adding g cyclic prefix samples The reference parameters of kmin and kmax are shown in the Table I The numerical values of the OFDM parameters are shown in the table II [1] 013 Engineering and Technology Publishing 57

2 argmax Journal of Communications Vol 8, o 9, September kmax j kn/ x( n) X ( k) e (1) kk min TABE I: CARRIER UMBERS FOR EACH MODE Symbol -1 symbol symbol +1 Robust Mode Carrier Spectrum occupancy patterns A K min K max B K min K max C K min K max D K min K max I ' I Figure 1 The transmitted symbols TABE II: UMERICA VAUES OF THE OFDM PARAMETERS Robust T u (ms) T g(ms) T s(ms) Mode A B C D where T s is the OFDM symbol period, T g is the cyclic prefix, T u is the useful part of OFDM symbol After passing over DRM channel with t paths, the receiver symbol: zn ( ) Z Summation () Summation () t 1 j n/, t0 t z ( n) ( h ( n) x ( n)) e w ( n) () where is the symbol index, is the frequency offset normalized to carrier spacing which can be divided into two parts: IFO and FFO f, h ( n) is the channel impulse response of the tth path, w ( n) is the contribution of the AWG The receiver symbol in the frequency domain can be expressed: where H t, Z ( k) H ( k ) X ( k ) W ( k) (3) t 1 j tk / t (4) H ( k) H ( k) e t0 t, ( k ): the channel frequency response of the tth path, t :the time-delay normalized to sample interval of the tth path, W ( k ): the AWG III THE COVETIOA FREQUECY SYCHROIZATIO FOR DRM The carrier frequency offset is usually divided into part and al part, and it can be calculated respectively With the influence of the al frequency offset, the accuracy of frequency offset estimation will greatly decrease The al frequency offset estimation can use the most commonly used algorithm: M (maximum likelihood) 1 f Figure The al frequency offset acquisition algorithm Define two sets: M I {,, g 1} (5) ' I g {,, 1} (6) where I is cyclic prefix of the symbol According to the features of cyclic prefix, it contains the same elements in ' I ds + d m = 0 w - jf E{ z(n) z (n+ m)} = ds e m, n I 0 others where d s and d denote the energy of the useful symbol w and Gaussian white noise The algorithm is shown in Fig (7) M arg max( ( ) - ( ) ) (8) n g -1 ( ) z( n) z ( n ) n n g -1 (9) 1 ( ) ( ( ) z n z ( n)) (10) n 013 Engineering and Technology Publishing 573

3 Journal of Communications Vol 8, o 9, September 013 SR where =, SR is Signal to oise Ratio, M is SR+1 the estimation of time-delay So, the al frequency offset is: 1 f ( M ) (11) j f n y (n) z (n) e (1) where y (n) is the receiver symbol which eliminates al frequency offset Making FFT (Fast Fourier Transform) operation on prefix, we can get Y (n) y (n) and removing the cyclic A The Integer Frequency Offset Estimation based on Time Pilots Define a TRC (time reference cell) as: ( ) /104 P( k) e j k k (13) where is a pilot boost factor and ( k) /104 denotes a predefined phase rotation of the pilot cell denotes the set of TRC indices The exact position and phase rotation of the TRCs are depicted in [1] The traditional frequency synchronization algorithm based on the time pilot exploits the correlation of the time pilots and receiver symbol [17] d F Y k d P k (14) k arg max{ ( ) ( ) } where () is modular arithmetic, is the modulus operation, F is the maximum allowable of, d is the trial value of conjugation, Y ( k d) P ( k) k () denotes the complex j k/ { H ( k) P( k) e (15) k W ( k) P ( k)} The algorithm is susceptible to multipath and noise The algorithm in [18] is based on the partial correlation and splits the correlation into the B blocks ( B T T : the allowed symbol time-delay error) a, a B arg max{ Y( k d) P ( k) } (16) m 1 k B df m The algorithms in [17] [18] are suitable for DAB (Digital Audio Broadcasting) system that utilizes the block-type pilot, while they are not suitable for DRM system because the time pilots are not uniformly distributed The frequency estimation algorithm in [19], implemented by TRC partitioning for DRM systems, is proposed The TRC partitioning scheme is used for weakening the effect of frequency-selective fading: int eger m / c arg max{ Y( k d) P ( k) } (17) where c d F m1 n1 kp mn, and / m are the numbers of pilot clusters and sub-groups in the mth cluster, respectively, Pmn, is the set of TRC indices, rounds the element to the nearest ote that the total number of sub-groups is t c m 1 m / [0] B The Integer Frequency Synchronization Algorithm based on the Frequency Pilots Define a FRC (frequency reference cell) as: j ( k) /104 Q( k) e k 1 (18) where is a pilot boost factor and ( k) /104 denotes a predefined phase rotation of the pilot cell 1 denotes the set of FRC indices The exact position and phase rotation of the FRCs are depicted in [1] One frequency synchronization algorithm for DRM system based on the frequency pilots is proposed by Communications Technology Institute of Darmstadt University The following will introduce the algorithm [1]: Firstly, the algorithm should make FFT operation on receiver symbols to estimate the power spectrum: () 1 aver 4s 1 j nk 4 ( ( ) ) s s (19) i 0 n 0 aver R k z n i e where s g, aver is the number of spectra used for averaging In the presence of frequency offset, the peak of the pilot will shift The frequency offset can be obtained by detecting the shift of the pilots g P ( k) 4 k ( 1) k 1 fac ^ fs f arg max R ( d Pfac( k)) 4s d k where fs is the sampling rate, frequency offset IV A The Proposed Algorithm ^ f THE PROPOSED AGORITHM (0) (1) is the estimation of Based on the characteristics of DRM frequency pilots, this paper proposes one frequency offset estimation algorithm for DRM The proposed algorithm utilizes M+1 consecutive received symbols After the conjugate multiplication of two adjacent symbols, the power spectrum of corresponding carriers Pow ( k) i : 013 Engineering and Technology Publishing 574

4 Journal of Communications Vol 8, o 9, September 013 Pow i ( k) Y i ( k) Y i1 ( k) k [0, 1] i [0, M 1] () where is the index of symbols According to (3), the phase Ph i( k) of Pow i( k) : Ph ( k) ac tan( Pow ( k)) k [0, 1] i [0, M 1] (3) i where i ac tan() is the function of phase acquired Define the angle vector Angl _ temp : M 1 Angl _ temp( k) cos( phi( k)) i0 M 1 (4) j sin( ph ( k)) i0 i where j 1 Therefore, frequency offset can be obtained by (5): int eger Angl _ temp( d k1) Angl _ temp( d k) arg max d Angl _ temp( d k3) d[0, 1] (5) where d is the trial value of, k1, k, k3 are the position of the frequency pilots without frequency offset Finally, combined with the al part, the actual frequency offset is B The algorithm analysis ˆ f (6) Before the estimation of frequency offset, the time-delay and al frequency offset have been already compensated The proposed algorithm exploits the known correlation of frequency pilots Pow ( k1 d) i Y k d Y k d i( 1 ) ( 1 ) ( H ( k1 d ) Q( k1 d ) i int eger int eger W ( k1 d) )( H ( k1 d ) i i1 int eger Q( k1 d ) W ( k1 d) ) Pow ( k d) i int eger Y k d Y k d i( ) ( ) ( H ( k d ) Q( k d ) i int eger int eger W ( k d) )( H ( k d ) i i1 int eger Q( k d ) W ( k d) ) Pow ( k3 d) i int eger Y k d Y k d i( 3 ) ( 3 ) ( H ( k3 d ) Q( k3 d ) i int eger int eger W ( k3 d) )( H ( k3 d ) i i1 int eger Q( k3 d ) W ( k3 d) ) int eger (7) (8) (9) If d, according to (18): int eger Q( k1 d ) Q ( k1 d ) int eger int eger j ( k1 d ) /104 e (30) j ( k1 d ) /104 e The Pow i( k1 d) can be expressed as : Pow ( k1 d) H ( k1 d ) i i int eger H ( k1 d ) int eger W( k1 d) W( k1 d) H ( k1 d ) i int eger (31) Q k d W k d ( 1 ) ( 1 ) int eger W k d W k d i( 1 ) ( 1 ) H ( k1 d ) int eger Q ( k1 d ) W ( k1 d) int eger i According to the DRM channel parameters: s max (3) T (33) where max is the maximum time-delay So the DRM channel is flat-fading in the frequency domain [] When the SR is high, the Pow i( k1 d) can be expressed as : Pow ( k1 d) H ( k1 d ) i i (34) int eger So the Ph ( k) at the position of the pilots can be expressed as : i ph ( ) 0 [ 1,, 3] i k d k k k k (35) Angl _ temp( k) (cos( Ph ( k)) cos( Ph ( k))) M1 (sin( Ph ( k)) sin( Ph ( k))) M 1 cos ( Ph ( k)) sin ( Ph ( k)) cos ( Ph M 1( k)) sin ( Ph M 1( k)) cos( Ph ( k))cos( Ph ( k)) 1 sin( Ph ( k))sin( Ph ( k) ) (36) 1 sin( Ph M 3( k))sin( Ph M 1( k)) sin( Ph M ( k))sin( Ph M 1( k)) M cos( Ph( k))cos( Ph 1( k)) sin( Ph( k))sin( Ph 1( k)) sin( Ph M 3( k))sin( Ph M 1( k)) sin( Ph ( k))sin( Ph ( k)) M M Engineering and Technology Publishing 575

5 the mean error of frequency offset estimation Energy Journal of Communications Vol 8, o 9, September 013 Finally, Angl _ temp( k) can be expressed as: Angl _ temp( k) M cos( Ph ( k) Ph ( k)) 1 cos( Ph ( k) Ph ( )) k (37) cos( Ph M 3( k) Ph M 1( k)) cos( Ph ( k) Ph ( k)) M M 1 M k [0, 1] The equation (36) (37) is monotone decreasing in the range [0, ] When Ph ( k) - Ph ( k) Ph ( k) - Ph ( k) 0 1 M M 1 the value of Angl _ temp( n ) is equal to, M Due to the correlation, Ph ( k) - Ph 1( k) Ph M ( k) - Ph M 1( k) of frequency pilots tend to 0 to its fullest extent, while Ph ( k) - Ph 1( k) Ph M ( k) - Ph M 1( k) of other carriers are much more than 0 The conventional algorithms mainly exploit the power factor of the pilots in the correlated calculation As the power factor of the pilots is inevitably disturbed by the noise and multipath, the correlated peak may be weakened The proposed algorithm gets rid of the power factor of the pilots in correlated calculation and only retains the phase factor of the pilots Because the DRM channel is flat-fading in the frequency domain, the phasedifference Ph ( k) - Ph 1( k) Ph M ( k) - Ph M 1( k) can efficiently weaken the impact of the noise and multipath In addition, superposition of pilots and a set of consecutive received symbols can be used to combat the noise Consider the complexity and accuracy of the proposed algorithm, 10 consecutive symbols are selected Dp 1 Hz 1 Hz Channel 5 Path 1 Path Path 3 Path 4 gain 1 1 delay 0 ms 4 ms Df 0 Hz 0 Hz Dp Hz Hz Channel 6 Path 1 Path Path 3 Path 4 gain delay 0 ms ms 4 ms 6 ms Df 0 Hz 1 Hz 4 Hz 36 Hz Dp 01 Hz 4 Hz 48 Hz 7 Hz Df is doppler shift; Dp is doppler spread Fig 3 shows the frequency offset estimation of the proposed algorithm in the channel 6 Set the frequency offset to Hz The peak is at position of the 6th sampling point, and the estimated frequency offset is Hz according to (4), (5), (6) Simulation test shows that the proposed algorithm meets the requirements of the DRM system the frequency offset estimate SR=10dB the sampling points Figure 3 Integer frequency offset estimation V SIMUATIO The channel bandwidth is 10Hz and the Robust mode B is selected in this paper The DRM channel parameters are shown in the Table III T s is 666 ms, T g is 533 ms, 7 /8 the carrier spacing is 46 Hz TABE III: THE CHAE PARAMETERS Channel 1 Path 1 Path Path 3 Path 4 gain 1 delay 0 ms Df 0 Hz Dp 0 Hz Channel Path 1 Path Path 3 Path 4 gain 1 05 delay 0 ms 1 ms Df 0 Hz 0 Hz Dp 0 Hz 01 Hz Channel 3 Path 1 Path Path 3 Path 4 gain delay 0 ms 07 ms 15 ms ms Df 01 Hz 0 Hz 05 Hz 1 Hz Dp 01 Hz 05 Hz 1 Hz Hz Channel 4 Path 1 Path Path 3 Path 4 gain 1 1 delay 0 ms ms Df 0 Hz 0 Hz Frequency offset Figure 4 Frequency offset estimation range Define mean error of Frequency Synchronization: M( ˆ ) = E[ ˆ ] (38) where is the actual frequency offset, ˆ is the estimation of frequency offset Fig4 shows the mean error of frequency offset estimation Set the range of the 013 Engineering and Technology Publishing 576

6 Journal of Communications Vol 8, o 9, September 013 frequency offset from to 1600 The SR is 10 db Simulation test shows that approximately within the range [-1500, 1500], the accuracy of the proposed algorithm can fulfill the requirements of the DRM system Define MSE (Mean Squared Error) of Frequency Synchronization: J( ˆ ) ˆ E (39) Channel the proposed algorithm Reference 19 Reference 1 Reference 18 Reference 17 minor changes for the MSE cure of the proposed algorithm with respect to Fig5, while the algorithms in [17], [18], [1] can no longer meet the requirements of frequency offset estimation with large MSE Compared with Fig5, the performance of the algorithm in [19] is about 6 db smaller As the fading caused by multipath channel makes the symbol power spectrum ups and downs, the accuracy of the algorithm in [1] can t be guaranteed The algorithms in [17], [18] must satisfy the condition that time pilots are uniformly distributed, are not suitable to the DRM system Fig 7 shows the MSE of the proposed algorithm versus SR in DRM channels Channel 1 is the AWG channel, while channel 6 is multi-path channel Compared with channel 6, the MSE of channel 1 is about 1dB smaller MSE channel 1 channel channel 3 channel 4 channel 5 channel SR/dB Figure 5 MSE of the traditional algorithms and the proposed algorithm versus SR in channel MSE Channel 6 MSE the proposed algorithm Reference Reference 1 Reference 18 Reference SR/dB Figure 6 MSE of the traditional algorithms and the proposed algorithm versus SR in channel 6 The mean square error (MSE) of normalized frequency offset is used to measure the performance of the algorithm Fig 5 shows that the MSE of the proposed algorithm and traditional algorithms versus SR in the channel Set the frequency offset to 1504 Hz From MSE curves, it is clear that the proposed algorithm has a small variance than traditional algorithms Moreover, under the same MSE conditions, the performance of the proposed algorithm is about -4 db higher than the traditional algorithms Fig 6 shows that the MSE of the proposed algorithm and traditional algorithms versus SR in the channel 6 Set the frequency offset to 04 Hz There are only SR/dB Figure 7 MSE of the proposed algorithm versus SR VI COCUSIO Based on the characteristic of the frequency pilots, one frequency offset estimation method is proposed for DRM system The proposed algorithm gets rid of the power factor of the pilots in correlated calculation and only retains the phase factor of the pilots The structure of the angle vector is beneficial to weaken the impact of the noise and multipath Simulation results show that the proposed algorithm improves the range and accuracy of the frequency offset estimation REFERECES [1] Digital Radio Mondiale (DRM)-System Specification ETSI ES V311, draft, Aug 009 [] M Morelli, A D'Andrea, and U Mengali, Frequency ambiguity resolution in OFDM systems, IEEE Commun ett, vol 4, pp , Apr 000 [3] C Chen and J i, Maximum likelihood method for frequency offset estimation of OFDM systems, Electronics etters, IEEE, vol 40, pp , Jun 004 [4] D Toumpakaris, J ee, and H- ou, Estimation of carrier frequency offset in OFDM systems based on the maximum likelihood principle, IEEE Trans Broadcasting, vol 55, pp , Mar Engineering and Technology Publishing 577

7 Journal of Communications Vol 8, o 9, September 013 [5] K Shi and E Serpedin, Coarse frame and carrier synchronization of OFDM systems: a new metric and comparison, IEEE Trans Wireless Commun, vol 3, pp , July 004 [6] M Henkel and W Schroer, Pilot based synchronization strategy for a coherent OFDM receiver, in Proc WCC, March 007, pp [7] J i, G iu, and G B Giannakis, Carrier frequency offset estimation for OFDM-based WAs, IEEE Signal Process ett, vol 8, pp 80 8, Mar 001 [8] S Attallah, Blind estimation of residual carrier offset in OFDM systems, IEEE Signal Process ett, vol 11, pp 16 19, Feb 004 [9] V Fischer and A Kurpiers, Frequency synchronization strategy for a PC-based DRM receiver, in Proc APCCS, Dec 004, pp [10] J-J van de Beek, M Sandell, and P O Borjesson, M estimation of time and fequency offset in OFDM systems, IEEE Transactions on Signal Processing, vol 45, pp , Jul 1997 [11] B G Yang, K B etaief, R S Cheng, et al, Timing recovery for OFDM transmission, IEEE Journal on Selected Areas in Communications, vol 18, pp 78-91, ov 000 [1] T M Schmidl and D C Cox, Robust frequency and timing synchronization for OFDM, IEEE Transactions on Communications, vol 45, pp , Aug 1997 [13] H Minn, V K Bhargava, and K B etaief, A robust timing and frequency synchronization for OFDM systems, IEEE Transactions on Wireless Communications, vol, pp 8-839, July 003 [14] B Park, C Hyunsoo, C Kang, and D Hong, A novel timing offset estimation method for OFDM systems, IEEE Communications etters, vol 7, pp 39-41, May 003 [15] F He, F Yang, C Zhang, and Z C Wang, Synchronization for TDS-OFDM over multipath fading channels, Consumer Electronics, IEEE Transactions on, vol 56, pp , 010 [16] X Y Jiang and Z W Dong, Carrier frequency offset estimation of DRM receiver, in Proc 4th International Conference on Proceedings Microwave and Millimeter Wave Technology, Aug 004, pp 8-85 [17] H ogami and T agashima, A frequency and timing period acquisition technique for OFDM systems, in Proc PIRMC, vol, Sep 1995, pp [18] K Bang, Cho, J Cho, H Jun, K Kim, H Park, and D Hong, A coarse frequency offset estimation in an OFDM system using the concept of the coherence phase bandwidth, IEEE Trans Commun, vol 49, pp , August 001 [19] Eu-Suk Shim, J B Kim, and Y-H You, ow-cost frequency offset estimation for OFDM-based DRM receiver, IEEE Trans Consumer Electronics, vol 56, pp , ov 010 [0] Y-H You and K-W Kwon, Multiplication-Free estimation of frequency offset for OFDM-Based DRM systems, Signal Processing etters, IEEE, vol 17, pp , Oct 010 [1] A F Kurpiers and V Fischer, Open-source implementation of a Digital Radio Mondiale (DRM) receiver, in Proc inth International Conference on HF Radio Systems and Techniques, June 003, pp [] K i and M Yan, Simulation of digital radio mondiale channel model, in Proc IEEE 3rd International Conference on Communication Software and etworks, May 011, pp Cheng Yan received his BS degree from the School of Electronic Information and Control Engineering at Shandong Polytechnic University Jinan, China, in 011 He is currently pursuing a MS degree at Communication University of China His research interests include signal processing, mobile multimedia and wireless communication Yan Ming received his BS in communication engineering from the anjing University of Posts and Telecommunications (JUPT) in 00 and his MS in signal and information processing from the Communication University of China (CUC) in 006 He received his PhD in communication and information system from CUC in 01 In January 01, Dr Yan Ming joined the GxSOC Research Institute at CUC as a Research Assistant Dr Yan Ming leads the China Mobil Multimedia Broadcasting (CMMB) Research Group at Communication University of China, whose mission is to conduct tracking research of mobile multimedia broadcast technology and the related support services in China He also participates with the Team for Research in broadband multimedia communication at the Institute of Microelectronics of the Chinese Academy of Sciences (IMECAS) In addition to his research activities, Dr Yan Ming serves as the Administrator of Postgraduate Studies in GxSOC Research Institute Dr Yan Ming is an Expert Member of both the ext Generation Broadcasting (GB) workgroup and the Audio Video Standard (AVS) workgroup 013 Engineering and Technology Publishing 578