This paper presents an error modeling of an odometry system for a synchronous-drive system and a possible strategy for evaluating this error. The odometry error is modeled by introducing four parameters characterizing its systematic and nonsystematic components (translational and rotational). The nonsystematic errors are expressed in terms of a covariance matrix, which depends on both the previous four parameters and the path followed by the mobile robot. In contrast to previous approaches which require assuming a particular path (straight or circular) in order to compute this covariance matrix, here general formulas are derived. We suggest a possible strategy for simultaneously estimating the four model parameters. As we will show, our strategy only requires measuring the change in the orientation and position between the initial and final configurations of the robot, related to suitable robot motions. In other words, it is unnecessary to know the actual path followed by the robot. We illustrate the proposed strategy by discussing the accuracy of the parameters estimation and by showing some experimental results obtained with the mobile robot Nomad150.