Model order reduction by Miller's theorem and root localisation
Model order reduction is a critical subject in the area of bottom-up behavioural modelling. A new method for model order reduction, which is the combination of classical Miller's theorem and the root localisation (RL) algorithm, is introduced. This method is based on frequency-independent Miller's theorem, RL and signal-path tracing algorithms. The first step of the presented procedure identifies the significant nodes in a circuit, discarding those deemed insignificant. Next, Miller's theorem, RL and a set of predefined simplification rules are applied, creating an order-reduced equivalent circuit. Such an approach maintains the capability of representing both the linear and nonlinear, the static and dynamic behaviours of the original circuit. This method can also be used to identify the topological nodes that have strong nonlinear dynamic behaviours. Finally, the extension of the simplified circuit to represent the nonlinear behaviour of the original circuits is presented.
