Fixed order anti-windup augmentation is addressed by considering the dynamics of the mismatch between the constrained and unconstrained responses. The method utilizes LMIs, and the result is similar to other recent LMI-based anti-windup synthesis methods. Robustness is directly addressed-which is not done in previous LMI work. An optimal LMI-based synthesis procedure is provided for static and plant-order linear anti-windup augmentation and the performance of the resulting design strategy is shown via a simulation example.