Design of a power optimized 1024-point 32-bit single precision FFT processor

Fourier transform is the basic operation between time and frequency domain transformation. As a key operation of digital signal processing system, Fast Fourier Transform (FFT) is widely used in many fields such as communication, biomedical signal processing and image processing, which require a high precision of processed signal. To meet the requirement, the floating point number can be used to improve the signal quantization noise ratio (SQNR) greatly. In this paper, we proposed a power optimized 1024-point high precision FFT processor with 32-bit single precision floating point number for both input and output. Simulation results show that the proposed FFT works at 1.2 V, consumes 17.6 mW with a 0.13 μm CMOS technology. Its SQNR can reach 97 dB, which is ~2× higher than the conventional design when maintaining the same power consumption. In the meanwhile, it gains ~99% power reduction when maintaining the same SQNR.