Rehabilitation Research Before Multicore DSP-based Real-time Wave

People have been using optical telescopes to conduct astronomical observations fora long time. Wavefront errors introduced by varied sources make it impossible toachieve theoretical resolution of a telescope thus decrease the quality of observedimages. In1953, Babcock tried to solve this using the concept of Adaptive Optics. Thecore idea of adaptive optics is to measure the wavefront errors and make compensationsto optical system in real time.As an important component of adaptive optical (AO) systems, wavefront controlleris used to process the measured error signal and output servo control signal in real time.Its processing latency and capacity directly influence the correction capability of an AOsystem. As we build larger modern telescopes, the computational complexity ofwavefront processing raises dramatically, especially the case with wavefrontreconstruction. The processing power of current wavefront controller is hardlyinsufficient.The research presented in this paper is aimed at finding an implementation ofwavefront reconstruction for AO systems with around one thousand servos. Theimplementation should have the following features: low latency, short developmentperiod, scalable, small in size. After analyzing the processing requirements ofcommonly used reconstruction algorithms, a multicore digital signal processor (DSP)C6678is chosen as processing hardware. A method consisting of four steps is used toget minimum latency. The steps are: parallelization of reconstruction algorithm, codeoptimization, optimization of execution sequence and decreasing data transfer time. Toaddress scalability of the implementation, the data interfaces of C6678is studied and asoftware design for setting parameters at runtime is proposed. Lastly, an experiment isconducted for a system with968sub-apertures and913actuators. With a single C6678chip, the final wavefront reconstruction implementation added merely57.2us to overallsystem latency.