Nonlinear Error Correction For Two-dimension PSD

When PSD (Position Sensitive Detector) applied, how to overcome the nonlinearproblem is an important research direction. In order to solve this critical problem, LMalgorithm which based on the improved BP neural network algorithm to advantagedthe PSD non-linear, namely use the neural network can approximate any nonlinearfunction to establish a non-linear mapping relations after neural network trained,achieved the PSD nonlinear compensation.Studied PSD linearity based on PSD structure works and the impact ofenvironmental factors, analyed double-time interpolation algorithm, biquadraticinterpolation algorithm, improved biquadratic interpolation algorithm, verified theadvantage of improved biquadratic interpolation algorithm by experiment, throghanalyed double time interpolation algorithm and biquadratic interpolation algorithmand gain their overcomes. Focus on study the BP neural network algorithm, obtainedthe lacks of BP neural network algorithm, analyed the BP neural network algorithmheuristic Improvement Act and the LM algorithm; established experimental platform,analyed using BP neural network algorithm and LM algorithm which based onimproved BP neural network algorithm to corrected PSD non-linear error; whentraining data, useing two hidden layer neural network training.Based on experiment, research useing interpolation algorithm, BP neural networkalgorithm and LM algorithm to corrected2-D PSD nonlinear error, according to thesimulation and simulation through MATLAB7.1contrast, results showed: LMalgorithm can effectively reduce the2-D PSD nonlinear to measurement results,compared to traditional BP neural network algorithm convergence rate, high accuracy,made2-D PSD linearity in nonlinear regions approximation close to linear aftercorrected.