| The embedded system is an integrated and complicated digital system. With the development of information technology, higher request to the performance, volume and power of embedded system has been put forward, which requires the design of embedded system (including software design and hardware design) to be optimized further. FFT is a basic algorithm in the field of digital signal processing, however, with the development of embedded system, especially the wide application of FPGA, we have raised the new demand on its time performance, space performance and precision. So how to design FFT algorithm at the embedded system platform ECNUX, enabling it to meet each need of embedded system, is the key content of this paper.Firstly , the properties of Fourier Transform, including discrete Fourier Transform, FFT and based-2 DIT FFT, are introduced in this paper; Secondly, ECNUX platform and ECNU-C language .mainly its characteristics and grammars we will use, are introduced briefly; Thirdly, in chapter four the situation of overflow of fixed point FFT is analyzed in detail ,at the same time two methods of controlling the overflow of fixed point FFT and the parallel scheme of FFT ,including parallel in butterfly and between butterfly, are provided; Finally, two parallel and fixed point FFT algorithms are simulated on platform VC6.0, and their time performance and precision are both compared by analyzing their results, which shows that the precision of the first method is better and the other's time performance is better.Because traditional FFT algorithm is realized in floating point, while ECNUX platform only supports fixed point operation, and FPGA is more suitable for fixed point operation, so the problem of realizing FFT in fixed point is solved in this paper at first. Secondly, ECNUX platform supports parallel operation, and FPGA has the ability of running in parallel, so the realization of FFT running in parallel is another key problem solved in this paper.
|
PDF Full Text Request