sally wrote:I thought you would go, even to explain the math in the time of the make timestrech, pitchshift, QLink realtime tune (for example) ... all mathematics for my 2500 please (xparis, you could talk in deep about all this please?)
why would you think we would talk about 'all math' in a thread about swing?
on older samplers, timestretch was accomplished by by breaking the sample into equal sized 'chunks', for example 1000 samples long, and crossfading between them while stepping through the sample length starting at different samples at a particular rate. the different 'presets' (percussive, vox1, vox2, etc) i THINK were simply adjusting the chunk size.
im not sure if the 2500 or the 1000 work this same way, but if you make a couple extreme examples i can hear, i could probably tell you.
the more modern way to do it is in effect very similar, in that you again take a part of the sample using various 'windows' similar, transform it via fft into frequency components, resample that to a new size, and the ifft it back into time-domain representation. then you put these back together. IF this is how the new machines do it then i'm sure the presets relate to window size/type. honestly i kind of doubt this is what they are doing because these phase-vocoders were not as popular back when these came out.
pitch shift works the exact same way in both cases you just resample (tune) the sample before the windowing/chopping.
finally, if you have been following along you would know that xparis is a legit dude working on the ren and i dont believe he is a programmer nor has he worked on the older mpcs. still, he might know or could ask someone else, but you'll probably have just as much luck reading some basic dsp books yourself instead of putting your finger in your nose.
thats as far as ill go because i have not experimented with the 2.5k or 1k, so im really just guessing, though those are the standard techniques for anyone really.
