![]() ![]() ![]() For sound this is as close to the perfect resampling method as is theoretically possible as the ear is incapable of being discriminative in these high bands (most of us can barely tell a 1kHz difference above 10kHz, and most, especially natural sources, sound like noise to us in those bands anyway).īit depth difference is of even lesser concern. A linear interpolation (the trivial up/down sampling solution) cuts off those frequencies very far from the Nyquist frequency, while leaving those close it almost unattenuated.Ī sinc filter, which is what is used by best interpolation methods, cuts sharply in the frequency domain, and removes, depending on sinc depth, almost all of the alias. The source of aliasing can be observed (I'll use imaging because it's easier to imagine) so much in the misalignment of the resolutions that causes say, a nice line to become a jagged mess.īut actually the issue is that in the frequency domain (where a smooth transition in an image, or a low tone in sound are in the lower band, and sharp transitions in image, or a high tone are in the high band) in these small differences most of the aliased frequencies are nearby the new Nyquist. Because it's done using deep sin(x)/x interpolation in any editor worth anything there is no such thing as 'suboptimal' conversion really (as opposed to, say, 88.2 -> 44.1 if that's what you're thinking), and that can be proven mathematically fairly easily but I'll try intuitively.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |