A major requirement in the analysis of multi‐input‐multi‐output digital networks is to compute the frequency response at a large number of points from a knowledge of the graph of a network with specified input and output nodes. Though the usual method of analysis by solving linear equations is more efficient than matrix inversion, solution of linear equations at each frequency may still require a large amount of computation. A more efficient method is to determine the poles and zeros of the desired system function and then calculate the frequency response from them. Though poles of the system function could be readily identified as contained in the eigenvalues of the state matrix of the network, difficulty is seen to arise in the determination of zeros which can not so easily be calculated. A simple and efficient method is proposed reducing the determination of zeros to a standard eigenvalue problem. Copyright © 1979 John Wiley & Sons, Ltd.