Synchronization And Channel Estimation For ACO-OFDM Wireless Optical Communications Based On Superimposed Training

The combination of wireless optical technology and OFDM (Orthogonal Frequency Division Multiplexing) is achieved perfectly by ACO-OFDM (Asymmetrically Clipped Optical-Orthogonal Frequency Division Multiplexing). Considering that only a quarter of subcarriers are used to transfer data in ACO-OFDM system, the system transmission efficiency is essential for ACO-OFDM system. But in practice training symbols and pilots are usually used to achieve timimg synchronization and channel estimation. Though satisfactory performance may be achieved, the system transmission efficiency must be seriously reduced. The superimposed training techniques make the training sequence and data transfer within the same period of time interval. Compared with the traditional time domain training symbols or frequency domain pilots methods, the superimposed training techniques increase the system spectral efficiency while ensuring the synchronization and channel estimation performance. Because the optical intensity modulation is be used in ACO-OFDM system, and the signals must be unipolar, non-negative and real. The kind of traditional superimposed training methods cannot be used directly.In this paper, the basic principles of ACO-OFDM and the transmission model of wireless optical communication system are introduced. The problems and reasons, which are encountered when some classical timing synchronization methods are used in ACO-OFDM system, are carefully analyzed. The necessity of using superimposed training techniques for timing synchronization in ACO-OFDM system is pointed out. In the traditional timing synchronization method using superimposed training, the training sequence is bipolar and complex. So it can not be directly used in ACO-OFDM system. A timing synchronization method for ACO-OFDM based superimposed training is proposed in this paper. Firstly, the Hermitian symmetry of IFFT (Inverse Fast Fourier Transform) is used to generate unipolar non-negative real sequence. Secondly, the correlation between training sequence and local sequence is used to realize timing synchronization, and the power allocation factor is used in process of decision threshold, that makes the decision threshold more scientific. Thirdly, the training sequence is separated with the data in time domain. Via simulation, compared with other methods, the proposed method has a significant improvement in synchronization accuracy, and estimator variance both in LOS and diffuse channel. In ACO-OFDM system, the mean of the data is not zero. That cannot satisfy the qualification in the traditional channel estimation method using superimposed periodic training, which is zero-mean of the data sequence. Via the improvement on the training sequence and the local matrix, a channel estimation method for ACO-OFDM based superimposed training is proposed in this paper. Firstly, the unipolar non-negative periodic real sequence is generated through splitting and composing complex sequence. Secondly, according to statistics of the received signal in time domain and the power allocation factor, the local matrix is designed rationally. Thirdly, the first-order statistic of the received signal in time domain is used to estimate the channel finite-impulse response. Fourthly, the correlation matching approach is used to estimate the unknown dc-offset. After simulation and theory reasoning, the effectiveness of the proposed method is tested, and its performance is analyzed.