Study On Methods To Process Deplacement Sensor Signal Of Fast Steering Mirror

Fast steering mirror (FSM) is an essential part of the tracking system, thestructure of FSM has a small stroke, high precision, fast response, and dynamic lagerror. In the fast steering mirror system, the measuring accuracy of the displacementsensor directly determines the accuracy of the system. Displacement sensor usuallyhas issues such as non-linearity and temperature drift, this part of the error has to becompensated; in addition, the random error of the circuit and signal quantization errorin the sampling process can’t be ignored, this part of the error manifests as signalnoise, which must be filtered. The reason why the fast steering mirror systempossesses high precision, a very important reason is the use of multi-sensor to obtainthe redundant displacement information, aiming at improving accuracy. The usualmethod is to filter the signal of each sensor separately and then utilize the linearoperation in order to mirror deflection information, but such treatment makesnon-filtered noise superimposed. On the contrary, it may increase the system error.Therefore, to gain a method which can effectively filter out the noise and take fulladvantage of the redundant information in the signal processing is very necessary.This paper launches a series of studies for the removal of random noise and theelimination of the redundant information.Firstly, The paper briefly introduces the structure character and central capabilityparameter research of the fast steering mirror, then analyzes the advantages anddisadvantages of different displacement sensors. Considering a variety of factors, thefinal scheme is the selection of the eddy current sensor as a fast mirror displacementsensor. After that, according to the characteristics of the eddy current sensor signal, itcompares the de-noising effect of different de-noising methods.This paper mainly studies the de-noising of the eddy current sensor randomsignal. After comparing several de-noising methods, the wavelet method is found toeffectively filter out random noise in the eddy current signal, prior to other methodsespecially at de-noising the single signal, in order to verify the real-time wavelet,itanalyses the amplitude-frequency characteristic of the wavelet. In addition, thepaper presents an improved adaptive independent component analysis method to establish a separate single-source model, the results show that, this method can notonly effectively eliminate the redundancy of the multi-sensor, but also has a goodde-noising effect. It builds a DSP-based control and algorithm processing platform,and utilizes the real-time analysis of the results of independent component analysisand wavelet de-noising. Results show that, the wavelet method has better real-timede-noising effect, while the independent component analysis method can effectivelyremove the redundancy. In addition, the algorithm of the independent componentanalysis method has a lager calculation and shows a poorer performance in real-timeaspects.