讨论了探测器的观测空间转换问题。介绍了一种用神经网络建立的从电信号特征参数到光路参数的光路反模型的方法。根据四像限光电探测器的两路输出信号在过零点附近时间段的逼近直线的斜率和截距与光电探测器的三自由度安装位置,以及探测器光敏面的离焦量的特殊关系,建立了称之为模型1的四像限光电探测器光路逆模型。同时以探测器特定时刻输出电压作为观测量,建立了称之为模型2的探测器光路逆模型。并以探测器光敏面的离焦量为例,给出了两个模型的实测值和模型1的重复性测试值。重复性测试值表明,模型1的最大重复测试误差只有0.015 mm。实测结果证明,模型1的检测精度可以达到微米级,而用探测器特定时刻输出电压建立的逆模型的检测精度只能达到毫米级,这证明用探测器输出信号过零点附近时间段的逼近直线参数作为观测信号优于用输出信号特定时刻电压作为观测信号。

In this paper, the problem about observation space of detector is discussed. A method of modeling the anti-models of ray-path from electric signal to parameter of ray-path using neural network is introduced. Making use of the special relations between the parameters of fitting lines at the crossover points of outputs of four-quadrant photodetecor and the three-dimensional free position, the distance between photosurface and focus of photodetector, the anti-model called model 1 of ray-path of four-quadrant photodetector is modeled. At the same time, a model called model 2 is modeled using the voltages of the output of photodetector at special time as observed quantities. Taking measuring for the distance between photosurface and focus of photodetector as an example, the measuring results are given, and the measuring results for repeatability of model 1 are also given. The measuring results for repeatability of model 1 show that the maximal measuring error is only 0.015 mm. And the measuring results show that the measuring accuracy of model 1 reaches micron level and that of model 2 reaches only millimeter level. Thus, that using the parameters of fitting lines at the crossover points of outputs of four-quadrant photodetecor as observation quantities is ahead of that using the voltages of outputs of photodetector at special time as observation quantities