Research On Random Error Suppression Technology Of MEMS Rate Gyroscope In Small Optoelectronic Equipment

The turntable is an indispensable part of the small optoelectronic equipment,and the gyroscope used for speed measurement is one of the main components of the turntable equipment.This topic intends to use the MEMS gyroscope for measurement,but the accuracy of the MEMS gyroscope is relatively low,which limits this Improve the overall measurement accuracy of photoelectric equipment.In view of this,this subject intends to study the random error suppression technology of MEMS gyroscopes to improve the accuracy of MEMS gyroscopes.The specific discussion is as follows:First of all,in order to analyze and suppress MEMS gyroscope noise,based on the existing experimental equipment,a MEMS gyroscope random error data acquisition system with FPGA-CPU as the core was built,and the random error data of the MEMS gyroscope was collected and compared with the data.The MEMS gyroscope has been calibrated and fundamentally analyzed.Secondly,in order to make up for the shortcomings that the Kalman processing is easy to build up by the model and the wavelet threshold denoising is not as effective as the multi-scale fusion method in suppressing some low-frequency noises,a Kalman array and real-time wavelet threshold joint filtering method is proposed.In the static and dynamic experiments,the results showed that the noise variance of the gyroscope was reduced by nearly 45 times under the static experiment,and the noise variance of the gyroscope was reduced by 62.3% under the dynamic experimentThirdly,the output data of MEMS gyroscope has weak nonlinearity.It is planned to apply the new empirical mode decomposition algorithm(CEEMDAN)to the noise reduction of MEMS gyroscope,but the application of CEEMDAN in real-time denoising is affected by the variance of the added noise.And the limit of the number of iterations.In order to solve this problem,a new non-stationary nonlinear filtering algorithm CEEMDAN-LWT that combines LWT and CEEMDAN is proposed,which not only solves the mixing problem,but also improves the uncertainty in the subfrequency..In the specific experimental results: in the static experiment,the gyroscope noise variance was reduced by nearly 14 times,and in the dynamic experiment,the gyroscope noise variance was reduced by 70.0%.Finally,in order to select the optimal algorithm to be implemented on FPGA,and to complete the comparative analysis of the algorithm on the software and hardware,the above-mentioned algorithm is compared in the frequency domain and the time domain.First,use the sweep signal to test the bandwidth of each filter to obtain the bandwidth of each filter;then,in the time domain,the noise signal is processed,and the noise variance of the filter and the filter time of the software algorithm are used as evaluation indicators.Comprehensive time domain and frequency domain experimental results can be obtained: Although the filtering effect of CEEMDAN-LWT is better,the filtering time of Kalman-real-time wavelet and Kalman array methods is shorter,and the bandwidth difference is not big.After completing the simulation implementation of Kalman array algorithm on FPGA,comparing the running time of the algorithm on FPGA(hardware)and MATLAB(software),it is found that the single-step algorithm on FPGA has lower latency,which is conducive to the real-time implementation of noise reduction algorithm.