In this work two observer based techniques are applied to OFC detection. The first approach uses a nonlinear observer with several adjustable gains, some of them are placed in the arguments of the nonlinear function corresponding to a given servo-loop model. The gains are analytically optimized using the monotone systems approach in order to maximize the observer robustness against external disturbances and sensitivity with respect to the fault signal. The second approach is based on High-order-sliding-mode (HOSM) differentiators and numerical optimization techniques, which ensure the faulty signal estimation in the presence of measurement noise and parametric uncertainty. Next, the estimated fault signal can be used to isolation or compensation. Simulation results confirm good fault detection abilities of both approaches.