This paper presents the design and real-time implementation of a composite controller consisting of a state-feedback control law and high-gain observer for LCL filter system in grid-tied applications. The main function of the feedback control law, designed in the $dq$ rotating reference frame, is to add a damping effect to the closed-loop system with a view to cancel the effect of the resonance frequency inherent in LCL filter system. High-gain observer is added to the feedback controller to achieve asymptotic regulation and disturbance rejection in the presence of model uncertainties and unknown disturbances. Specifically, high-gain observer is considered as a disturbance observer to compensate for the effect of parametric uncertainties and unknown inputs. High-gain observer has a rate of convergence that is dependent on its gain, which is required to be large enough to ensure fast disturbance estimation. However, high observer gain can lead to poor steady-state performances due to the effect of measurement noise. This concern is overcome in this work by using an existing adaptive technique to adjust the observer gain on-line. Moreover, the resulting high-gain observer is further simplified to provide a less-complex estimator and to show its integral action property. The performance of the composite controller was experimentally tested. The obtained results demonstrated the ability of the proposed controller to meet the control objectives with a reduced effect of measurement noise.