| Sensor is the core component in the measuring systems, and its accuracy play a decisive role in the performance of measuring systems. Now, the distortion of input signal caused by the dynamic characteristics of system frequently happens. The key to enhance the performance of sensors lies in correcting or compensating their dynamic characteristics, which has already become an important subject researched both at home and abroad at present time.This dissertation analyzes several kinds of dynamic compensation methods:the system identification method, Pole-zero cancellation method and the method based on neural network. The influence of noise on dynamic compensation of system identification and the influence of modeling error on dynamic compensation of pole-zero cancellation is studied. The relationship between modeling error of the system and widening multiple of the working frequency band is analyzed. And the relationship between the change of noise after compensating and widening multiple of the working frequency band is analyzed quantitatively. In this dissertation, the compensation principle of neural network is introduced. The appropriate model of dynamic compensation and input signal are chosen. The impact of the change of noise on the neural network after compensation is studied. And the relationship between the size of noise in the input signal and modeling error of dynamic compensation is analyzed quantitatively. Simulation results show that these methods can be used to complete sensors dynamic compensation.Digital signal processing experiment platform is used in the paper, which takes TMS320F2812 processor as the core. On the basis of digital signal processing evaluation board, a real-time dynamic compensation system is developed for sensors. The sensor is simulated by mimic channel. The real-time dynamic compensation system under different transfer functions of sensor model is developed. Results show that this system can complete the expectant function. |
PDF Full Text Request