- Sep 6, 2013Hello Thorsten,

you wrote:So, am I right in saying that each of the 16384 columns in the exported text file is a frequency bin of 1.346Hz width, and column dB1 is a range of amplitudes (over time) starting at 1.346Hz, column dB2 is a range of amplitudes starting at 2.692Hz, and so on… (with column dB0 starting at 0Hz).If I’m right, this may be the closest I can get to producing a range of amplitudes for individual frequencies, which I can then divide by the amount of amplitudes to get an average. It seems to make sense to me.

The interpretation is almost perfect... maybe with an error (offset) of a half FFT bin width. Quite academic if you don't really need this degree of frequency resolution, but here is the reason:

The 'first bin' has 'DC' (0 Hz) in the *center*. In fact, if the FFT input was complex, the 'DC' bin would cover frequencies from -half_bin_width to +half_bin_width. Here, since the FFT input (in the time domain) is real, negative frequencies fold back into positive frequencies, but the border between the 'first' and 'second' bin is still at +half_bin_width.

If the exported FFT file shall contain a header line, the coloums titled "BW_Hz" contains the width of each frequency bin in Hertz, and the column titled "F1_Hz" contains the *center* frequency of the first bin, also in Hertz. If the FFT input is real (not complex), this column contains zero in each line, because that's the *center* frequency of that bin - not the *start* frequency.

All the best,

Wolf .

- << Previous post in topic Next post in topic >>