This letter introduces several techniques which allow one to create high-quality models of time domain signals produced by the finite-difference time-domain method. Guidelines for automated selection of key model parameters, such as the length of training sequence for model building and the model order, are given. The application of the criteria for extraction of time signature features based on the initial time sequences proposed in the paper is illustrated on two microwave filters. Index Terms-Finite difference-time domain (FD-TD), signal processing.