From: Ana Celia Vila Verde (
Date: Mon Nov 23 2015 - 09:19:35 CST

Dear VMD users,

I recently compared the results of the spectral densities calculated
using the specden package in vmd (with normalization fourier), using the
fftpack also available in VMD, and also using Matlab, using a very
simple input timeseries: the sum of two sinusoidal waves plus noise. In
Matlab notation: x=sin(2*pi*50*t) + sin(2*pi*120*t) + 2*randn(size(t)).

I found that specden identifies only one of the frequencies of the
timeseries, whereas both fftpack and matlab correctly identify both
frequencies. This problem only occurs for particular timeseries,
though: for example, when I increase the amplitude of the sinusoidal
waves (e.g., to x=10*sin(2*pi*50*t) + 10*sin(2*pi*120*t) +
2*randn(size(t))), then specden also identifies 2 frequencies. Even in
this case, though, the relative intensities of the two peaks differ from
those calculated by either fftpack or matlab.

I thought this information might be useful to other VMD users.