Convergence design considerations of low order Q-ILC for closed loop systems, implemented on a high precision wafer stage

Iterative learning control (ILC) design is quite often based on the principle of using the tracking error of a system to obtain an ideal input to reduce the tracking error of the system. For a closed loop system this may not be an optimal choice. It is shown that in a closed loop setting, using the tracking error signal for ILC may result in conflicting feedback and feedforward actions. This can be avoided by adding frequency weighting to the measured errors but in the paper it is shown that the ILC can be designed without any extra filters by considering what the desired ILC behavior should be and choosing the input signal accordingly. The measured results of two quadratic criterion ILC designs are presented, only differing in the choice of the input for the ILC, yet still yielding considerably different results. The designs have been implemented on a high precision wafer stage with direct links to modern wafer steppers used in IC manufacture. The results from these experiments are used to illustrate the influence of the design choice of the input signal for the ILC.