Efficient near optimal maximum likelihood symbol timing recovery in digital modems

Maximum likelihood (ML) estimation theory provides a general framework for developing optimum timing recovery schemes for digital communication systems. Feedforward schemes based on ML-theory have been shown to have rapid acquisition characteristics. Another important trend is to use digital interpolation for symbol timing adjustment. The so-called Farrow (1988) structure has become popular in this context because it offers an efficient and flexible realization structure for such an interpolation task. In a previous paper Vesma and Tuukkanen (see Proc. IEEE NORSIG 96, Espoo, Finland, p.183-6, 1997) proposed a new feedforward timing estimation scheme which utilizes the Farrow structure efficiently. The scheme is based on a low-order polynomial approximation for the log-likelihood function and also the derivative of this function is available with a small additional computational cost. The performance of the new timing recovery technique was found to be close to ideal even with short symbol block-lengths which is very important in mobile systems. In this paper, we show that this new scheme gives much better results when optimized training sequences are used. The scheme works also with severe multipath distortion channel.