Abstract
Ïn this paper, the authors have established the design criteria by which mean-squared-error value is minimized. Though they are rough, we can apply them effectively for the design of compensator.
When noise is present, Wiener's theory of Optimum Filter is applied, and when noise is negligiblly small, the saturation of the controlled process is taken into consideration, using Booton's linearization.
If the bandwidth of control system is sufficiently larger than that of input signal and noise is small, it is shown that our criteria agree with those which minimize the integrated squared-error value for step input. Some examples of design by those criteria are also shown.
