A stationary process subjected to a linear periodic time-varying filter becomes cyclostationary. Under certain inversibility conditions, perfect linear reconstruction of a bandlimited input process can be obtained when the filter is known. In this paper, we show an original implementation method to realize a linear periodic filter by means of periodic clock changes (PCCs) that corresponds to a physical model and presents a low implementation cost. We also present how periodic clock changes can be implemented in practice. Moreover, we introduce a method to estimate the parameters of the obtained filter necessary for the reconstruction. This estimation is particularly useful when the signals are sampled. We apply these results to analog scrambling. Examples are given both for a continuous valued signal and a binary signal.