EEG signal is a kind of very complex non stationary signal. It is of important clinical values for using EEG most efficiently to distinguish all kinds of components in EEG or getting needed components to judge some pathology features effectively. The phase moving of the system can change the relative phase relation between the components in the inputed signal, so there can be a great change of the feature in time domain of inputed signal. The FRR and RRF digital filtering methods and their realization are discussed. By using these methods, the system can be of the feature of zero phase error, and the filtering of EEG signals with zero phase error can be realized. Based on the definition of sleep spindles, we can detect sleep spindles in time domain directly by using digital filtering with zero phase error.