The control algorithms of electrical machine drives rely on models that are classically valid for signals in low-frequency ranges versus the current sampling frequency and the range of eddy currents. Some position self-sensing methods, which are used in motion-sensorless controls, are based on the injection of high-frequency signals, for which the model needs to be refined, taking the discretization and the eddy-current effects into account. In this paper, we introduce different circuit models for ac permanent-magnet machines, which are valid from low- to high-frequency spectra, for the purpose of analysis. Combined with the identification of two experimental machines, the goal is to highlight some conclusions about the suitable frequencies for self-sensing.