DFT, DFT and Fourier Series.
Cuthbert Nyack
The application of both the Fourier Series and the DFT result in
discrete spectra. The Fourier Series uses a continuous periodic function
in the time domain and gives a nonperiodic discrete spectrum.
The discrete periodic function used by the DFT results in a
discrete periodic spectrum. The difference between the 2 is in the periodicity of the spectrum. There is also a numerical
difference because the Fourier series uses areas and an averaging
factor, while the DFT uses samples and no averaging factor.
The applet below illustrates the relation between the 2. If the
sampling frequency is high enough so there is no aliasing, then the DFT spectrum between 0 and ws/2 multiplied by the "averaging factor" 2/Ns (Ns is the number of samples) is equal to the
Fourier Series. Increasing Nf2 to 100 show the usual periodicity
of the DFT spectra, while the Fourier Series spectra remains
nonperiodic.
The eg parameters (3.0, 4.0, 0.01, 11.39, 40, NA, 100, 150.0, 26.0, 4.0)
show that aliasing has started to affect accuracy of the equivalence.
Reducing the sampling frequency by reducing Ns introduces more
aliasing. Leaving Ns at 40 and changing T2 to 12.26 removes the
spectral leakage and restores the equivalence.
For the Fourier Series, the lines in the spectrum are given by
wn = n2p/T rad/s where T is the period.
For the DFT the lines are given by
wm = m2p/NT rad/s where T is the sampling time
and N is the number of samples.
Return to main page
Return to page index
COPYRIGHT © 2007 Cuthbert Nyack.