On checking causality of bandlimited sampled frequency responses

We propose a new technique for checking causality and self-consistency of bandlimited sampled frequency responses. The main algorithm is based on an efficient numerical implementation of the Generalized Hilbert transform. The key advantage of the proposed formulation is the explicit derivation of error bounds for the unavoidable sources of inaccuracy due to the finite number of available frequency samples. These bounds are used for detecting any significative causality violations in the data under analysis. The proposed scheme can be applied in CAD environment to certify the quality of frequency data coming either from measurement or simulation, before using the data in the actual design flow.