This gives you the signal as Y(f) vs f where f is the frequency (which you would normally plot on the X axis). Take an FFT Power spectrum of this sample. You have a signal (a temperature in this case) that is oscillating in some fashion - possibly not sinusoidally - and you want to track the frequency of this oscillation over a period of time (which one assumes must be longer than the period of the osciallation) to look, for example, for a drift in the frequency of the oscillation.Īssuming this is correct, what you need to do is to take a sample of your data that is longer than the expected oscillation period but shorter than the time over which you expect the frequency to change. Ok, let me try to see if I'm understanding.