The problem of the design of a perfect reduced-order unknown-input observer for standard systems is formulated and solved. The procedure of designing the observer using well-known canonical form is proposed and illustrated with a numerical example. Necessary and sufficient conditions for the solvability of the procedure are given.