Fast Fourier Transform - Digital Signal Processing and Applications with the C6713 and C6416 DSK - page 237

FIGURE 6.21. Time-domain plot of the radix-4 FFT magnitude of a 2-kHz input sinusoidal
signal using FFTr4.c .
as in Example 6.5. This includes the FFT function as well as the function for gen-
erating the digit reversal index and the digit reversal function.
Build this project as FFTr4 . Input a 2-kHz sinusoidal signal with an approximate
amplitude of 2 V p-p. Verify the output in Figure 6.21. These results are similar to
those obtained with the radix-2 FFT function in Example 6.4 and the radix-2
C-coded FFT function in Example 6.2.
A project application in Chapter 10 makes use of the real-time radix-4 FFT func-
tion with frequency-domain filtering.
6.8.1 Fast Convolution
The following examples show how the FFT enables signals to be processed in the
frequency domain. Fast convolution [19,20] takes less computational effort and is
potentially more accurate than time-domain implementation of FIR filters having
very large numbers of coefficients.
Example 6.7: Fast Convolution with Overlap-Add for FIR Implementation
Using TI's Floating-Point FFT Functions ( fastconvo )
Figure 6.22 shows a listing of the program fastconvo.c to implement an FIR filter
and illustrate the fast convolution's overlap-add scheme [19,20]. TI's floating-point
FFT support functions, bitrev , digitrev_index , and cfftr2_dit were