Procedures for Determining the Parameters of the First-Order Canonical Models of Class A and Class B Electromagnetic Interference [10]

Simple approximate methods as well as various more precise, analytical procedures for determining the first-order parameters of both the approximate and exact Class A and B noise models are derived and illustrated for both the ideal case of infinite sample data and the practical cases of finite data samples. It is shown that all first-order parameters of these models can, in principle, be obtained, exactly or approximately, from the ideal or practical measurements. (All first-order parameters of Class A but only the (first-order) even moments of the Class B models are exactly obtainable in the ideal cases.) Procedures for establishing meaningful measures of the accuracy of the parameter estimates in the practical cases are also identified: these include suitably adjusted, nonparametric, small-sample tests of " goodness-of-fit" (such as Kolmogorov-Smirnov tests), which provide the principal techniques for establishing accuracy, at an appropriately selected significance level (¿0). Questions of robustness and stability of the models, parameters, and their estimators are also discussed.