For closed-loop adaptive optics systems limited by time delay and meas
urement noise, we demonstrate that the ideal rejection transfer functi
on is proportional to the frequency signal-to-noise ratio of the wave-
front input. We describe a new modal linear predictive controller that
approaches this ideal transfer function. Its parameters are optimized
by minimization of the residual wave-front error with a modified recu
rsive least-squares algorithm. in the case of evolving turbulent condi
tions. improvements brought by the predictor. The optimization can be
performed with closed-loop data We present numerical simulations to sh
ow the significant (C) 1998 Optical Society of America.