Sensors are important components of the dynamic positioning system, take compass for instance, which measures the course of a ship. Based on the Finite Impulse Response (FIR) filter, a simple and efficient sensor fault diagnosis scheme was proposed to improve the ability for judging whether sensor is out work or not. This method improved the capability of diagnosing when and what faults occur by using double criterions of fault detection method, and solved the sensor fault diagnosis problem in output mutation. The double criterions consist of state deviation and signal-to-noise ratio (SNR). Only the output signal of both state deviation and SNR changed suddenly can prove a fault of sensor has occurred. If the output signal of state deviation transforms but the SNR does not, it just indicates that the sensor is disturbed by the external environment. Applying this method, an explicit solution was obtained, and achieved a greater capacity and efficiency in the implementation process. The validity and excellent performance of this algorithm was proved by the simulation results.