It is shown that the state transition matrix may be identified using arbitrary sampling under not very restrictive conditions. A good transmission of the relative measuring errors towards the identification results may be achieved by an adequate choice of the sampling instants using very well known numerical analysis concepts.