This paper presents a method to reduce the complexity of a linear or linearized (small-signal) analog circuit. The reduction technique, based on quality-error ranking, can be used as a standard reduction engine that ensures the validity of the resulting network model in a specific (set of) design point(s) within a given frequency range and a given magnitude and phase error. It can also be used as an analysis engine to extract symbolic expressions for poles and zeroes. The reduction technique is driven by analysis of the signal flow graph associated with the network model. Experimental results show the effectiveness of the approach.
Walter Daems, Georges G. E. Gielen, Willy M. C. Sa