FFT 1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20
1.1 Review of DFT The DFT of a(n), n =0, 1, ---, N —1, is defined as N-1 soar X(k) = SY n(nje Bnew n=0 N-1 = Yo(nWH, k=0,1,--,N-1, n=0 where Wr = e2/N x(n) can be reconstructed from X(k) by 1 Nel omni a(n) = = YS X(b)elanh k=0 1X , = x XW n=0,Le,] k=0 all X(k)/N is the complex amplitude of the sinusoidal component at frequency 2x/N of a(n). X(k) is called the frequency spectrum of 2(n). 1.2 Fast Fourier Transform (FFT) N-1 ; , X(k) = 5 (nye, WS n= Calculating one point of X(k) needs V complex multiplications (4 real mult.). Hence computing X(k), k=0, 1, +++, N—1 requires N? complex multiplications. -j2nk/N (4) FFT is a fast algorithm which requires | loa (5) complex multiplications only (radix 2). Major idea in FFT: 1. Break down a long sequence into small segments, 2. Compute DFT on small segments, 3. Combine DFT of small segments to form the final DFT output. E.g. If we decompose a length NV sequence to two sequences of length N/2, the DFT complexity is iN\2 N? vs. (5) x 2+ combine. (6) Comparison of DFT and FFT Time (sec. N DFT FFT 4N°/40MIPS 2NlogN/40MIPS DFT/FFT (f, = 8kH2) AN? 2NlogoN (sec.) (sec.) 0.082 256 = 262144 4096 0.0066 0.0000102 64 0.128 1024 4194304 2.480 0.1 0.000512 204.8 0.256 2048 16777216 45056 0.42 0.0011264 372.36

You might also like