Optimal fixedl-point fast fourier transform

FFT is an important IP for OFDM-type wireless transceivers. In the context of software defined radio, we consider implementing a scalable FFT IP that can easily be adapted to different FFT lengths while optimized for power and speed. In particular, we propose to develop word-length adaptable, fixed point butterfly units that can change internal word lengths to provide flexible trade-off between performance and power consumption. Given the FFT length and desired SQNR, we derived analytical solution to determine the desired word length for intermediate results as well as for representing parameters. Compared to previous results, our design yields better SQNR while using shorter word length.