In this paper, a new systematic approach to the design of feedforward-feedback (FF-FB) control systems is proposed, applicable in cases where bounds on plant and disturbance uncertainties are known. The robust performance of this control scheme is compared with that of two alternative feedback designs, on the basis of a test case in which the superiority of pure feedforward control is guaranteed for perfect models. On the basis of the results, the limits to the advantages of a combined feedforward-feedback controller over the optimal feedback compensator are determined as a function of plant and disturbance uncertainties. Design guidelines are laid for both FB and FB-FF designs where the significant disturbance can be measured and where model and disturbance uncertainties can be quantified