I have 25,200 samples of data. My bandwidth is 12.6KHz and my Fs is 1.26MHz, I want to plot an Amplitude-Frequency Spectrum to display up to 100 different signals, that's on purpose (12,600 * 100 = 1,260,000), For that I need to compute FFT on the first 100 samples, I can't really do that because the algorithm I'm using is a Radix-2 algorithm, I thought about padding with 0s to 128 but then my BW is changing to 9.74KHz. the other thing I thought is to compute FFT of 128 and just cut the last 28 samples, but then I can't see anything after 974KHz, so I think I have to use some other type of FFT algorithm, which one should I use? I want to implement one myself and efficiency doesn't really matter here, because I'm working with small amount of samples.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
You can use the Prime-factor FFT algorithm recursively, since $100 = 5 \cdot 5 \cdot 2 \cdot 2$, you only have to implement the radix-5 and radix-2 reductions.
-
$\begingroup$ is it one algorithm or 2 separate functions etc one is Radix-2 and the other one Radix-5? $\endgroup$ Oct 5, 2021 at 17:15
-
2$\begingroup$ There is one algorithm to reduce a FFT of size p * N into p FFTs of size N and then combine the FFTs in one single array. Read carefully the Wikipedia Algorithm, also check this out numericalrecipes.wordpress.com/2009/05/29/… $\endgroup$– BobOct 5, 2021 at 17:30