Bias analysis of continuous-time model identification from filtered sample output data

The purpose of this paper is to present some issues that arise when applying a traditional continuous-time approach to system identification. In particular, we consider the bias that is introduced into the continuous-time model through the use of sampled data which is filtered to recover the derivative of the continuous time signals. We find that this error has two constituent components, one being related to the sample size, as expected, the other related to the sample period. The results are confirmed by simulation examples.