Ron N. wrote: ...> But the problem is that the questions asked of the results > of FFT operators are often underconstrained. The questions > here are often not about the spectrum of just the window given > to the FFT, with perhaps discontinuities at the end-points, but > the spectra of some phenomena which has an existance > outside the window given to the FFT. In those cases, one might > look for the most probable result of N+M unknowns (spectra > produced by the data plus some unknown but assumed more > locally "continuous" or otherwise behaved data just beyond the > window) given only N data points inside the FFT aperature.A problem to be sure, but the mathematics of a Fourier transform constrains the number of meaningful results to the number of meaningful data. It neither prevents one from guessing nor from draping those guesses in window-dressing. Given the meaningful results, anyone can dress them as he pleases without adding to what the transform shows.> That's usually the case when you assume that some interpolated > sinusoid is more interesting than the odds that something > generated that same result using a whole distribution of only > integral frequency oscillators that just happened to look like > an aliasing.<metaphor mode> Those considerations can shape the dressmaker's dummy for the added dressing, but too often it justifies conclusions made of whole cloth. </metaphor mode> Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
FFTW question about larger output array than input
Started by ●July 21, 2006
Reply by ●July 25, 20062006-07-25
