【摘要】 为了解决光学膜厚监控系统中经锁相放大输出的监控信号精度较低、极值点附近变化不灵敏等问题,采用卡尔曼滤波(KF)对光学膜厚监控信号进行处理,建立了针对非线性光学监控系统的扩展KF模型,模型输出包括监控信号和导数信号。仿真分析首先利用TFCalc软件生成4层增透膜理想监控曲线,加入高斯噪声后,信噪比为15dB,通过卡尔曼滤波处理后监控信号信噪比改善达16dB且实时动态跟踪特性良好,延时小于1 s;利用平均值滤波处理KF输出导数信号,提前预判读获取导数过零点对应膜厚。同时通过对比TFCalc理想监控曲线极值点坐标和预判导数信号过零点坐标,得到理论膜厚误差小于4nm。此项研究提高了膜厚检测的准确性。更多还原

【Abstract】 In optical monitor system(OMS) for thin film deposition,the film thickness is insensitive to the monitor signal when the signal is near the peak.The accuracy of the monitor system is limited by the noise of the signal.A generalized Kalman filter(KF) was applied to process the monitor signal.Based on nonlinearity of OMS,mathematical model and arithmetic of extend KF was constituted.An ideal monitor curve of 4-layer AR film was generated by TFCalc and mixed with a Gaussian noise.As a result,the signal-to-noise improvement of monitor signal after KF reached 16dB with excellent dynamic tracking characteristic and the delay time was less than 1s.By comparing the theoretic peak coordinates and zero coordinates of first derivative after KF,the calculated thickness error was less than 4nm.The study improves the accuracy of the film thickness detection.更多还原