Identification of continuous time SISO differential equation models-noise-free case

This paper looks at the problem of identifying continuous time transfer functions from samples of input/output data. We suggest the use of continuous time Laguerre filters as state variable filters and the use of higher order Pade approximations to give discrete time approximations of these continuous time filters. This gives rise to a numerically well-conditioned algorithm as the discrete time filters so obtained are shown to be approximately orthogonal. The result is an accurate identification scheme as demonstrated by an example which shows that accurate parameter estimates can be obtained even at low sampling rate