In avionics and aerospace multisensor systems, reliable and early detection of individual sensor faults present substantial challenges to health monitoring designers of such systems. This study addresses the problem of sensor fault diagnosis. The proposed solution is based on a non-homogeneous high-order sliding mode observer used to estimate the faults, theoretically in finite time and in the presence of bounded disturbances. The sensor faults are estimated for the class of systems satisfying the structural property of strong observability. A key feature of the proposed solution is concerned by the effect that measurement noise could have on fault reconstruction. It is shown that the fault estimation error is bounded in the L∞-norm sense, and an upper bound is theoretically derived. The method is applied to the problem of sensor fault estimation of a large transport aircraft. Simulation results as well as a pilot experiment are presented to demonstrate the potential of the proposed method.