基于焦面光强的分块镜共相位误差校正

发布时间：2018-02-06 00:18

本文关键词： 分块镜共相位误差自适应光学无波前传感器随机并行梯度下降遗传算法调制传递函数 piston误差　出处：《北京理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：分块镜共相位误差检测方法主要分为两大类,一类是光学检测校正方法,包括迈克尔逊干涉法、窄带/宽带夏克-哈特曼法、色散条纹法、相位恢复法、相位差法等,这类算法都是通过光学干涉或者衍射方法对分块镜成像后的图像进行分析得到误差信息,并进行校正的。这些方法都需要改变原有光路或增加新的光学元件,如滤光片、色散棱镜等,增加了光学系统的复杂性;另一类分块镜共相位误差校正方法是电学方法,该类方法主要是在分块镜背部安装电容式、电感式或光纤式传感器,用于测量相邻子镜间的相对位移量,进而得到分块镜之间的拼接误差并进行校正,这类方法无法测量相邻子镜沿光轴方向的绝对距离,一般需要用光学方法事先标定出各个子镜间的绝对零位。本文以归一化的像清晰度函数作为评价函数,仿真分析了单色光和白光照明两种情况下分块镜共相位误差与评价函数的变化关系。仿真表明当采用单色光照明时,分块镜平移(piston)误差校正的动态范围小于0.5λ;采用白光照明时评价函数存在局部极值,必须采用全局最优化算法。当piston误差大于0.5λ时,采用随机并行梯度下降(SPGD)算法易陷入局部极值,而采用遗传算法可以搜索到全局最优解,但迭代速度较慢。本文提出基于无波前传感器自适应光学混合优化算法,该算法是结合随机并行梯度下降(SPGD)算法和遗传算法的优点,对分块镜倾斜误差小于0λ/D,piston误差在[-λ,λ]范围内的分块镜共相位误差进行校正。该混合优化算法先利用遗传算法的全局寻优能力在[-λ,λ]范围内搜索最优解的大致范围,后利用SPGD算法的快速寻优能力在[-0.5λ,0.5λ]小范围内迅速找到最优解,当piston误差在[-λ,λ]范围内时,混合优化算法的校正精度可优于10nm。为了进一步扩大分块镜piston误差的校正范围,本文提出建立起宽光谱调制传递函数(MTF)与分块镜piston误差之间的函数关系,以两块分块镜光瞳为仿真模型,通过建立的函数关系式对分块镜piston误差进行校正,仿真研究表明不同带宽(Δλ)下,分块镜piston误差的校正范围和精度也不同。当Δλ=1nm,piston误差在[-300λ,300λ]范围时,校正精度优于15λ;当Δλ=10nm,piston误差在[-60λ,60λ]范围时,校正精度优于8λ;当Δλ=50nm,piston误差在[-12λ,12λ]范围时,校正精度优于1λ,因此实现了分块镜大piston误差的检测和校正。针对更多块的分块镜光瞳模型,仿真分析了固定带宽下四种不同分块镜模型对应的MTF与分块镜piston误差之间的关系,仿真结果表明仅在两块分块镜相对排列时各分块镜间不能进行独立的标定和校正。
[Abstract]:The common phase error detection methods of block mirror are mainly divided into two categories. One is optical detection and correction methods, including Michelson interferometry, narrowband / wideband Shack-Hartmann method, dispersion fringe method, phase recovery method. The phase difference method is used to obtain the error information by optical interference or diffraction. These methods need to change the original optical path or add new optical elements, such as filter, dispersion prism and so on, which increase the complexity of optical system. The other method is the electrical method, which is mainly to install capacitive, inductive or optical fiber sensors on the back of the block mirror to measure the relative displacement between adjacent sub-mirrors. Then the splicing error between the sub-mirrors is obtained and corrected. This method can not measure the absolute distance of the adjacent sub-mirror along the optical axis. It is generally necessary to calibrate the absolute zero position between each sub-mirror by optical method in advance. In this paper, the normalized image definition function is used as the evaluation function. The relationship between the co-phase error and the evaluation function is analyzed under the condition of monochromatic light and white light illumination. The simulation results show that when monochromatic illumination is used. The dynamic range of error correction is less than 0.5 位; When white lighting is used, the evaluation function has local extremum, so the global optimization algorithm must be adopted. When the piston error is greater than 0. 5 位. The stochastic parallel gradient descent (SPGD) algorithm is easy to fall into the local extremum, while the genetic algorithm can be used to search the global optimal solution. In this paper, an adaptive optical hybrid optimization algorithm based on wavefront sensor is proposed, which combines the advantages of stochastic parallel gradient descent SPGDalgorithm and genetic algorithm. The tilting error of the split mirror is less than 0 位 / D Piston error. [The common phase error of the block mirror in the range of-位, 位] is corrected. The hybrid optimization algorithm first uses the global optimization ability of genetic algorithm. [Search for the approximate range of the optimal solution in the range of-位, 位, and then use the fast optimization ability of the SPGD algorithm. [The optimal solution is found quickly in a small range when the piston error is within the range of 0. 5 位 / 0. 5 位. [In the range of-位, 位, the correction accuracy of the hybrid optimization algorithm is better than that of 10 nm. In order to further expand the correction range of the piston error of the split-mirror. In this paper, a functional relationship between the wide-spectrum modulation transfer function (MTF) and the piston error of the split-mirror is proposed. The simulation model is based on the pupil of two block mirrors. The piston error of block mirror is corrected by the established function relation. The simulation results show that there are different bandwidth (螖位). The correction range and accuracy of the piston error of the block mirror are also different. [In the range of -300 位 / 300 位, the correction accuracy is better than 15 位; when 螖位 = 10nmPiston error. [The correction accuracy is better than 8 位 in the range of -60 位 ~ (60 位). [In the range of -12 位 ~ (12 位), the correction accuracy is better than 1 位, so the detection and correction of the large piston error of block mirror are realized. The relationship between the MTF and the piston error of the four different split-mirror models with fixed bandwidth is simulated and analyzed. The simulation results show that there is no independent calibration and correction between the two chunking mirrors when they are arranged relative to each other.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】