FFT peak estimation - binning

Hi,

I want to determine a tendency of the dominant peak in the frequency domain of a dataset. I use 6 seconds of data and sample it at 20Hz. I calculate the FFT and I can determine the dominant peak. But now I want to get a better view and I want to bin the FFT function and determine the value of the biggest peak. I want bins with a size of 0.1Hz. Does anybody knows how to do it, because the method I'm using now takes sometimes 2 values from the original FFT for 1 bin and sometimes just 1 value from the original FFT for a bin and this gives a very bad result.

Thank you in advance,

Peter

Answer 1

[bym]

Unless I am missing something, you need a longer sample than 6 seconds. Frequency resolution = FFT points/fs

In your case 120/20 = .1667 To get .1 Hz resolution you would need 200 points or 10 seconds of data

[Walter Roberson]

Thud! Guess I wasn't quite awake yet when I did my calculation :(

[bym]

Never thought I'd get the drop on the rascally raccoon!

[Walter Roberson]

Or to put it a different way, for any fixed sampling frequency, provided the Nyquist criteria is met, frequency resolution = 1/T, so to get 0.1 Hz resolution, you would need 1/0.1 Hz = 10 seconds of data.