There are a number of procedures of analysing the zig-zag test to obtain the steering quality parameters of a relevant ship, ranging from the simple method with the first-order (K and T) equation to the phase-plane analysis taking the higher-order time constants and non-linearity into account. Growing importance of closed-loop studies of steering control and increasing number of less course-stable ships (e. g. very large tankers) call for the latter type of comprehensive mathematical model of steering response. However, it is not an easy task, unlike the simple K-T analysis, to define all the parameters involved in such mathematical model from a zig-zag test record. The more numbers of parameters to be defined, the harder to keep the numerical accuracy and to assure the feasibility of such analysis. The present paper relates to a new approach to this task which proved promising. That is, (1) to adopt numerical filtering to minimize the noise involved in the yaw-rate record and at the same time to obtain a reasonably noise-free yaw-acceleration ; the result gives us a measured yaw-acceleration to yaw-rate phase-plane trajectory of a relevant zig-zag test. (2) to adjust the parameters of the mathematical model built in an analogue computer by turning a number of knobs, keeping one eye to the phase-plane trajectory displayed on a cathode-ray oscilloscope, so that the displayed trajectory coincide with the measured one. The mathematical model employed is the rudder-to-yaw response equation with a cubic-type non-linear term, i. e. T ₁ T ₂ψ+ ( T ₁+ T ₂) ψ+ψ+αψ³= K δ+ K T ₃δ