Parametric reduction of Jacobian matrix for fault analysis

The paper deals with a fast numerical method for generating the Jacobian matrix of network function for large analog linear circuits. It is based on the use of cofactor matrices. The computational cost of the Jacobian is comparable to the cost of computing a single network function on selected frequencies. In addition, the method allows performing the reduction of Jacobian matrix based on large-change sensitivities in order to decrease the computational complexity of subsequent testability analysis. An example analysis of a frequency filter is presented.