In this work a new method for designing predictive controllers for linear SISO systems is presented. It uses only one prediction of the process output J time intervals ahead to compute the correspondent future error. The predictive feedback controller is defined by introducing a filter that weights the last w-predicted errors. In this way, the resulting control action is computed by observing the system future behavior and also by weighting present and past errors. This last feature improves the closed-loop performance to disturbance rejection as shown through simulation results.