We consider the problem of constructing functional filters (optimal functional observers, i.e., observers for linear functionals of the state vector) for linear time-invariant control systems in which the inhomogeneity contains additive white noise as a term in addition to control. The output of the system is linear in the state vector and also contains additive white noise as a term. With the help of canonical representations, a comparative analysis of the second- and third-order filters by the mean square observation error in the steady state is carried out. An example of a fourth-order system is given, showing that with an increase in the dynamic order of the filter, the optimality by a quadratic criterion increases.