A mathematical operation that analyses an arbitrary waveform into its constituent sinusoids (of different frequencies and amplitudes). This relationship is stated as
where
s(
t) is the waveform to be decomposed into a sum of sinusoids,
S(
f) is the Fourier transform of
s(
t), and
i = √−1. An analogous formula gives
s(
t) in terms of
S(
f), but with a normalizing factor, 1/2π. Sometimes, for symmetry, the normalizing factor is split between the two relations.
The Fourier transform pair, s(t) and S(f), has to be modified before it is amenable to computation on a digital computer. This modified pair, called the discrete Fourier transform (DFT) must approximate as closely as possible the continuous Fourier transform. The continuous time function is approximated by N samples at time intervals T:
The continuous Fourier transform is also approximated by
N samples at frequency intervals 1/
NT:
Since the
N values of time and frequency are related by the continuous Fourier transform, then a discrete relationship can be derived: