The effect of identifier structure on parameter convergence

The effects of a chosen system structure on the identification of its parameters are considered. In particular, convergence rates are compared for the parameters of three state-space structures: direct form II, parallel, and dual generalized Hessenberg representation structures. It is shown that the chosen structure does influence identification algorithms, and that this influence is measured by examining the information contained in the structural parameters. A conjecture about the relative convergence of the parameters is offered, and evidence in its support is provided. An important result is that identification of the usually identified direct from II parameters (the standard ARMA parameters) does not necessarily yield the fastest parameter convergence for the system being identified.<>