In this report we study estimation of time-delays in linear dynamical systems with additive noise. Estimating time-delays is a common engineering problem, e.g. in automatic control, system identification and signal processing. The purpose with this work is to test and evaluate a certain class of methods for time-delay estimation, especially with automatic control applications in mind. The class of methods consists of estimating the time-delay as a continuous parameter with a prediction error method in some simple model structures which are often used in process industry. The methods are evaluated experimentally with the aid of simulations and plots of RMS error, bias, standard deviation and confidence intervals for different cases.The results are: It is best not to prewhite the input signal. There should be at the most one real pole in the model structure. In some cases the simplest model structure, a first order system with time-delay "idproc6", is clearly the best. It is not clearly worse than the best in any case. The RMS error varies much with the system, the input signal type and the SNR. For idproc6 it varies between 0.3 and 12.1 sampling intervals in the performed simulations with a mean of 4.7.