| The Discrete Fourier Transform (DFT) is one of the most widely used algorithms in Signal-Processing and Communication Systems. Traditionally, digital-design techniques have been used to implement DFT through Fast Fourier Transform (FFT). The arithmetic nature of this algorithm, involving just scaling and addition, lends itself to an analog/mixed-signal implementation which offers a better solution in some specific applications.;In this work, a quantitative model to analyze performance of current mode analog FFT is developed. The impact of twiddle-factor precision and mismatch on FFT performance is analyzed quantitatively with mathematical models. The performance of different popular FFT implementations is evaluated under these conditions. A possible use of this analog FFT in a BPSK receiver is also explored. This is followed by a circuit level implementation of current-mirror based FFT. It is demonstrated that for small N-point, the performance of analog FFT matches closely with the traditional implementation. |
