With the increasing complexity of today’s spacecrafts, there exists a concern that the on-board flight computer may be overburdened with various processing tasks. Currently available processors used by NASA are struggling to meet the requirements of scientific experiments [1, 2]. A new computational platform will soon be needed to contend with the increasing demands of future space missions. Recently developed hybrid field-programmable gate arrays (FPGA) offer the versatility of running diverse software applications on embedded processors while at the same time taking advantage of reconfigurable hardware resources, all on the same chip package. These tightly coupled HW/SW systems consume less power than general-purpose singleboard computers (SBC) and promise breakthrough performance previously impossible with traditional processors and reconfigurable devices. This thesis takes an existing floating-point intensive data processing algorithm, used for on-board spacecraft Fourier transform infrared (FTIR) spectrometry, ports it into the embedded PowerPC 405 (PPC405) processor, and evaluates system performance after applying different hardware and software optimizations and architectural configurations of the hybrid-FPGA. The hardware optimizations include Xilinx’s floating-point unit (FPU) for efficient single-precision floating-point calculations and a dedicated single-precision dot-product co-processor assembled from basic floating-point operator cores. The software optimizations include utilizing a non-ANSI single-precision math library as well as IBM’s PowerPC performance libraries recompiled for double-precision arithmetic only. The outcome of this thesis is a fully functional, optimized FTIR spectrometry algorithm implemented on a hybrid-FPGA. The computational and power performance of this system is evaluated and compared to a general-purpose SBC currently used for spacecraft data processing. Suggestions for future work, including a dual-processor concept, are given.
Library of Congress Subject Headings
Fourier transform infrared spectroscopy--Data processing--Evaluation; Field programmable gate arrays; Computer algorithms--Evaluation
Department, Program, or Center
Computer Engineering (KGCOE)
Blavier, Jean Francois
Bekker, Dmitriy, "Hardware and software optimization of fourier transform infrared spectrometry on hybrid-FPGAs" (2007). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus