Energy-Efficient Fast Fourier Transform for Real-Valued Applications
Charalampos Eleftheriadis, Georgios Karakonstantis
Abstract
This brief presents a new energy efficient Fast- Fourier Transform (FFT) architecture for real-valued applications. The proposed architecture decimates the FFT in time domain with bit-reversed inputs which allows to avoid the use of all costly complex FFTs operations required by the existing schemes. This leads to the reduction of the required memory by a factor of 2 while processing two inputs in parallel, thus doubling the throughput and improving the energy efficiency compared to the current real-valued FFT designs. Furthermore, the output frequencies are computed at their natural order by using a novel memory management technique, without requiring any reordering circuit unlike existing works. In summary for a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> point FFT the proposed architecture leads to an increased throughput of 2 samples per clock cycle, requiring <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$N-2$ </tex-math></inline-formula> memory cells, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$8logN-8$ </tex-math></inline-formula> real adders and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$3logN-4$ </tex-math></inline-formula> real multipliers. Our results show that we can achieve up to 46.86% energy savings when compared with recent real-valued FFT architectures.