Two Maximum-Likelihood Symbol Synchronizers with Superior Tracking Performance

In this paper we consider a data-aided and a non-data-aided synchronizer, which are both based upon the maximum-likelihood estimation principle. For both synchronizers, the linearized mean-square synchronization error, the mean acquisition time, and the acquisition range are evaluated in the case of NRZ-PAM. We demonstrate that the linearized mean-square synchronization error is very close to the theoretical lower bound, and about 10 dB smaller than for a conventional synchronizer with the same closed-loop bandwidth. This superior tracking performance is confirmed by computer simulations. Comparing the acquisition times on the basis of equal acquisition ranges, we find that the non-data-aided ML synchronizer is about a factor of 2 faster than the conventional synchronizer. For the data-aided ML synchronizer, the result of the comparison depends highly on the amount of frequency detuning.