Research On High Performance And Real-time Computing For Adaptive Optics

Adaptive optics system is a key instrument in large ground-base telescope, laser beam cleanup system and laser beam transfer system. Adaptive optics real-time wave-front processor is the primary component of adaptive optics system, whose performance is vital to the residual error after correction using adaptive optics system. This dissertation is dedicated to the research on how to realize the real-time processor based on multi-core CPU and the optimisation reconstruction method for lase beam cleanup system on this platform. The content of this dissertation is organized as following 4 sections:1. The computational performance requirements for the AO system with large num-ber of actuators is first studied. Influence of the delay caused by the real-time controller on the error after correction using an adaptive optics system with an integration controller in Kolmogorove turbulence is analyzed, and the results show that the objective of design of realtime controller is to realize the high performance computation in a very short du-ration. The contribution to the error from the delay caused by the real-time controller in an actual adaptive optics system is analyzed, and comparison results with the classic Greenwood error estimation model show that the error is very close to the results from Greenwood estimation if the fc in Greenwood model is the rejection bandwidth at -3 db.2. Performance and architecture of the real-time processor based on multi-core CPU is analyzed, especially for the most computation-intensive process-reconstruction pro-cess. A simple criterion is concluded to classify the performance and the CPU's suitabil-ity for the reconstruction. In order to assemble the real-time tasks and monitoring tasks in a single computer, an architecture using a real-time operating system is proposed and the performance is analyzed. To our known, it is for the first to realize a real-time processor for AO system over 2000 actuators operating over 2000 Hz with only one board using this architecture.3. The coupling problem of the tip/tilt mirror and deformable mirror is analyzed, and the result shows that stability is deteriorated using the least square algorithm by adding limitative aberration vector in close-loop AO system, so as the tip/tilt on deformable mir-ror can not be restrained. A project-constraint algorithm is proposed for this decoupling problem. Comparison result by simulation shows that the proposed method has better performance correcting for high-order aberration when tip/tilt error exists, and a better decoupling performance could be achieved using the proposed method. Meanwhile, the temporal error is very close to the error using the least square algorithm by adding limita-tive aberration vector when the algorithm is implemented on the real-time platform based on Multi-core CPU.4. Characteristic of slab laser aberration is analyzed, and results show that least-squares reconstruction method is not the optimal reconstruction method to achieve high-est intensity improvement in the far field for this non-uniform distribution of aberration. Difference in objective of optical transfer system and the least-squares reconstruction method is analyzed, and a weighted least-squares method is proposed to improve the beam quality after correction with adaptive optics system. The comparison results show that beam quality and concentration of energy is improved using the weight least-squares method in this slab laser system.