AbstractThe ‘neural’ A/D converter of Tank and Hopfield is analysed from the viewpoint of circuit theory. It is shown to be a particular case of the canonical non‐linear programming circuit of Chua and Lin. Several qualitative properties concerning the DC solutions are proved. A piecewise‐linear analysis of trajectories in the state space is carried out on the basis of the eigenvalues and eigenvectors of the corresponding dynamical matrix. It is shown that the eigenvalues are related to the driving point impedance of an RC one‐port network. Asymptotic expressions of the eigenvalues are derived for large N, N being the number of bits. It is shown that the dynamic behaviour represents a sort of successive‐approximation A/D conversion. However the conversion time does not increase appreciably as the number of bits increases. the cases of incorrect behaviour and the amount of the maximum conversion error are derived for N > 5. Simulation results are presented for N ≤ 5.