DSP: Spectrum with FFTW

[hush-hush]

Hello,

I currently begin to study DSP, so my first step is to get the spectrum of a song.

I use FFmpeg to decode the file, and FFTW to do discrete fournier transform and get the spectrum.

My probleme is that I can't get the upper part of my spectrum (the hight frenquency).

here is my code:
This function is call with sample in data, data contain 2 channel.
So n_data the number of data in each channel, and total_size is equal to n_data * 2.
"update_spectrum" update the display.
And rdft_data_in/rdft_data_out/plan are static, the function is always call with the same len of data.


static int calc_audio(float *data, int n_data, int total_size)
{
int i;

if (!rdft_data_in)
{
rdft_data_in = fftw_malloc(sizeof(*rdft_data_in) * n_data);
rdft_data_out = (fftw_complex*) fftw_malloc(sizeof(fftw_complex) * n_data);
plan = fftw_plan_dft_r2c_1d(n_data, rdft_data_in, rdft_data_out,
FFTW_MEASURE | FFTW_DESTROY_INPUT);
}

for(i = 0; i < total_size; i += 2)
rdft_data_in[i / 2] = (data[i] + data[i + 1]) * .5f;
fftw_execute(plan);

double *freq_data = calloc(n_data, sizeof(*freq_data));
for (i = 0; i < n_data; i++) {
float r1 = rdft_data_out[i][0];
float i1 = rdft_data_out[i][1];

freq_data[i] = sqrt(sqrt(r1 * r1 + i1 * i1) / sqrt(n_data));
}

update_spectrum(freq_data, n_data);
return 0;
}
here is the song with audacity (we can see that the file is a mp3, the upper frequency are null)
http://hush-hush.me/spectrum_auda.png

Here my result as you see it miss my the half of my spectrum and I can't understand why,
http://hush-hush.me/spectrum_test.png


Thanks for your help !