On Nov 19, 5:37 am, "Henik Vargas" <hvarg...@gmail.com> wrote: > > What makes you think you need a correction factor? > > Instead of replicating what matlab does, read up on > > the method you play with. > > > Rune > > Rune, > > The motivation for the correction factor comes from the fact that windowing reduces the power of the signal, therefore, the power of the psd is much lower than that of the unwindowed signal. Another way of looking at it is that the correction factor makes the power spectrum estimate asymptotically unbiased (See Hayes, Statistical Digital Signal Processing and Modeling for a reference). > > Also, I read the documentation for the function, and I think my question is a fair one. I could just blindly use the function as is, but then I lose a learning opportunity.
My point is that in the applications where spectrum estimation methods like Welch's method are used, the main focus is to get an *impression* of the spectrum shape. The exact numbers are irrelevant.