In this paper the problem of noise suppression for a magnetic levitation system is addressed. The problem is formulated as a nonlinear regulation problem; a state feedback and an error feedback internal model-based regulators are designed, able to offset the noise in spite of the presence of unknown parameters affecting the model of the system. In both cases a saturated control law is designed in order to fulfill physically constraint characterizing the system. Moreover a general result in the context of output regulation of linear systems with constant uncertain parameters affecting the inputs channel and with input constraints is addressed.