Christian Schoenebeck wrote: > Not dangerous at all. And we also though about that as well in the past. But > frequency domain != frequency domain. There are unlimited ways to store > samples in the frequency domain. For example the DFT/FFT uses equal spaced > (linear) bands, which is not the best solution for human perception. I.e. the > resolution in the lower frequencies is too bad and it wastes too much data > for the upper ones. And last but not least does the human ear prefer certain > frequency bands. So there are various other transformations which deal with > this issue (e.g. the Mel scale filterbank).
Yeah, it did occur to me afterward that, really, if you need to do frequency-domain processing, you can always set up an FFT yourself. We were talking about modules that would do that in the past. > Josh Green even proposed a vector based storage (in the time domain) of > samples a while ago. Not sure if this would be beneficial for something > though. That sounds interesting, like the difference between vector and bit-mapped fonts. I can imagine it would make for a good way to do high-quality pitch and time scaling, but there are other methods to accomplish that with standard samples. -- Darren ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Linuxsampler-devel mailing list Linuxsampler-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxsampler-devel
