I'm writing a very simple in-place DFT. I am using the formula shown here: http://en.wikipedia.org/wiki/Discrete_Fourier_transform#Definition along with Euler's formula to avoid having to use a complex number class just for this. So far I have this:
private void fft(double[] data)
{
double[] real = new double[256];
double[] imag = new double[256];
double pi_div_128 = -1 * Math.PI / 128;
for (int k = 0; k < 256; k++)
{
for (int n = 0; n < 256; n++)
{
real[k] += data[k] * Math.Cos(pi_div_128 * k * n);
imag[k] += data[k] * Math.Sin(pi_div_128 * k * n);
}
data[k] = Math.Sqrt(real[k] * real[k] + imag[k] * imag[k]);
}
}
But the Math.Cos and Math.Sin terms eventually go both positive and negative, so as I'm adding those terms multiplied with data[k], they cancel out and I just get some obscenely small value. I see how it is happening, but I can't make sense of how my code is perhaps mis-representing the mathematics. Any help is appreciated. FYI, I do have to write my own, I realize I can get off-the shelf FFT's.