On the techniques of clock extraction and oversampling

Based on a probabilistic analysis, we compare and contrast the characteristics of two well known techniques for data recovery in communication links based on serial data streams: clock extraction and oversampling. The approach chosen is that of determining the probability of error in recognizing the transmitted symbols in the presence of additive noise and sampling time uncertainty or jitter. Based on a simple signal model it is shown that the probability of error for a single sample and as a function of the sampling time point is a functional involving the probability densities of additive noise and time uncertainty and the symbol shape function. The resulting formulations lend themselves to an evaluation using numerical quadrature. Furthermore, the analysis can be used for a prediction of the bit error rates (BER) that may be expected in the clock extraction and oversampling techniques. The motivation for this study is to advance the understanding of the two approaches in this specific aspect of data capture robustness.