Continuous-time noise modeling from sampled data

Most stochastic processes in engineering applications have an intrinsic continuous-time (CT) nature, for example, the noise generated by resistors and semiconductor devices. A few methods exist that identify CT noise models from sampled data. In this paper, a new method based on the concept of filtered band-limited white noise is introduced, an in-depth analysis of the basic assumptions made by the different approaches is given, and the pros and cons of each method are discussed. Finally, the new method is illustrated on a real measurement problem, i.e., the operational modal analysis of a bridge.