The sensorless control schemes based on machine saliency detection by signal injection commonly adopt a position observer to estimate the motor position. The position observer usually employs proportional-integral-derivative- (PID-) type controllers and low-pass-filters (LPFs) whose parameters need to be properly tuned to achieve satisfactory, or at least stable, performances. Generally, the position observers are tuned with trial and error procedures that require commissioning time and control design experience. This paper deals with the observer modelling and tuning. In particular, the observer model is used for a model-based position control tuning and the effectiveness of the proposed procedure is confirmed by the good agreement of theoretical and experimental results. The experimental test have been realized using both a linear tubular permanent magnet motor (LTPM) and a rotating internal permanent magnet motor (IPM) to demonstrate that the motor control performances can be predicted only if the observer behaviour is considered.