The comparison between recursive least squares (RLS) and Kalman filter (KF) is presented in this paper, both methods were adequate to estimate six parameters of a synchronous machine. The work focused on finding the operating conditions which the quality of the identification achieved with Kalman filter is better than recursive least squares. A linear model of the machine is used in order to considerate the currents and their derivatives as the system inputs while the three-phase voltage signals are the outputs. Furthermore two experiments with simulated and measured data were carried out, three operating scenarios and two variations of the algorithms respectively were considered. Despite the great similarity and good performance of both methods, it was found that Kalman filter slightly exceeded least squares due to the fact that it presented smaller oscillations in the estimated value of the parameters for any operating condition.