为了准确利用远场得到近场相位分布, 提出了多帧相位反演算法。这是一种利用多个远场以实现传统Gerchberg-Saxton(G-S)算法的相位反演方法, 其中的新远场是通过叠加已知像差到待测像差后产生的。在此算法的基础上, 提出以变形镜面形来实现反演的方法, 并通过数值仿真和实验验证了这种基于变形镜面形的多帧G-S相位反演方法的可行性。仿真结果同时还表明, 采用4个变形镜面形产生相应的远场, 平均仅需50次的迭代便可反演出不同D/r0数值的大气像差, 这些反演的像差与其对应待测像差之间的差别的均方根值平均小于0.005 λ。

In order to obtain the near-field phase distribution by using the far field intensity information, a multi-frame algorithm is proposed. It is a phase-retrieval algorithm which carries out the original Gerchberg-Saxton (G-S) algorithm by using more than one far field, and the new far fields are generated by adding prescient aberrations to the unknown phase. A novel method to achieve the multi-frame G-S algorithm by using the deformable mirror (DM) is proposed, and its feasibility is certified by numerical simulations and experiments. The simulated results further imply that the multi-frame G-S algorithm is capable of recovering the atomspheric aberrations within 50 iteration loops, on condition that 4 DM surfaces are utilized to build the new far fields, and the root mean square average of residual errors between recovered phases and true phases is less than 0.005 λ, with λ as the light wavelength.