A computer algorithm for causal spectral factorization

The authors summarize the theoretical basis for a numerical algorithm for causal spectral factorization of a scalar frequency response by sampling and interpolation of the frequency spectrum. Matrix spectral factorization based on a Newton-Raphson recursion proposed by J.H. Davis and R.G. Dickinson (1983) is considered, along with alternatives for causal projection. Details of the development and testing of a computer code which implements the algorithm in a reliable and numerically stable way are provided, including options for enhancing efficiency of the spectral factorization for application to computation of optimal control laws for a class of transfer function models. Computational examples are given, including control of continuum models of flexible structures. The use of the algorithm for design of feedback control for such models is outlined. The implementation of real-time control by special-purpose computer architectures is dealt with.<
>