抄録
In this paper, we shall consider a lower bound of sampling periods, when continuous systems are identified by least squares method using discrete model. The lower bound is minimum sampling period such that the identification accuracy based on the step response does not exceed an allowable value. This work is motivated by the fact that too small sampling periods lead to the deterioration of identification results in the presence of noise, which is caused by the bias of the least squares estimates.
We investigate the effect of the bias on the identification accuracy, since the bias can be decided by N/S ratio and sampling period, and discribe graphically the relation between N/S ratio and sampling period for certain values of step response error in the case of the first, second and third order continuous systems. It is shown from this relation that the lower bound becomes larger, as system order increase. Using the figures, the lower bound can be found.
