基于多核DSP的自适应光学系统波前处理器的研究

发布时间：2018-05-14 10:39

本文选题：自适应光学 + 波前处理器　；参考：《中国科学院研究生院（西安光学精密机械研究所）》2013年硕士论文

【摘要】：自适应光学系统能有效克服大气湍流带来的波前误差。然而，大气变化的随机性和瞬时性给自适应光学系统的设计带来了很大的困难，其中波前处理必须完成的大量实时运算是一个难以克服的问题。当前波前处理器大多采用DSP（数字信号处理器）来实现波前处理运算，DSP处理能赋予系统高精度和高灵活性；但是传统DSP运算速度制约了波前处理器的性能，给系统带来了较大时延。本文研究了波前处理器需要完成的波前斜率计算，波前复原和控制运算等任务，，对Zernike模式法的各阶有效性进行了仿真，并且得出波前处理器的运算量。结合波前处理器算法流程和运算量需求，提出了矩阵分割，任务间流水线处理、子孔径行与子孔径内并行计算的并行处理策略；在此理论基础上本文设计了以FPGA和四个TMS320C6678为处理核心的并行波前处理平台的软件和硬件实现方案；该平台的理论峰值处理速度达到每秒1280GMAC，系统时延约为30us，大大提升了波前处理器的性能。结合平台设计方案，本文对高性能波前处理平台进行了硬件设计。该平台采用TI高性能8核DSP TMS320C6678为运算单元，四个TMS320C6678组成一个流水线运算阵列。硬件设计中本文采用分立结构实现系统电源设计，通过监控芯片完成上电顺序控制；同时设计了丰富的内外部高速接口和存储空间；为了满足多种图像格式输入的需要，另设计了宽度足够的自定义通用图像处理接口；文章最后对硬件平台做了PCB设计和信号完整性分析，并对关键信号进行仿真，验证了硬件设计的有效性。
[Abstract]:The adaptive optical system can effectively overcome the wavefront error caused by atmospheric turbulence. However, the randomness and instantaneity of atmospheric changes bring great difficulties to the design of adaptive optical systems, among which a large number of real-time operations that must be completed in wavefront processing are difficult to overcome. At present, most wavefront processors use DSP (Digital signal processor) to implement wavefront processing. DSPs can give high precision and high flexibility to the system, but traditional DSP speed restricts the performance of wavefront processor. It brings great delay to the system. In this paper, we study the tasks of wave front slope calculation, wave front recovery and control operation, and simulate the efficiency of Zernike mode method, and get the computation amount of wavefront processor. Considering the flow of wavefront processor algorithm and the requirement of computation, the parallel processing strategy of matrix partition, inter-task pipeline processing, parallel computation of sub-aperture line and sub-aperture is proposed. On the basis of this theory, the software and hardware implementation of parallel wavefront processing platform based on FPGA and four TMS320C6678 are designed. The theoretical peak processing speed of the platform reaches 1280GMAC / s and the system delay is about 30us. which greatly improves the performance of the wavefront processor. Combined with the platform design, this paper designs the hardware of the high performance wavefront processing platform. The platform uses TI high-performance 8 core DSP TMS320C6678 as computing unit and four TMS320C6678 to form a pipelined operation array. In the hardware design, the system power supply is designed with discrete structure, and the power supply sequence is controlled by the monitoring chip. At the same time, the rich internal and external high-speed interface and storage space are designed. In order to meet the needs of the input of various image formats, At last, the PCB design and signal integrity analysis of the hardware platform are done, and the key signals are simulated to verify the validity of the hardware design.
【学位授予单位】：中国科学院研究生院（西安光学精密机械研究所）
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP332

【参考文献】