Quote:
Originally Posted by innesireinar
too high SR the converters suffer during the reconstruction process
|
Not so much to do with reconstruction but if you sample zillion times a second you just have too much data to process and because ALL DATA HAVE "X" PCT ERRORS the more data you sample the more errors you will sample as well.
And because reconstruction takes all data into account, it also takes those bogus samples as well, which means if you only need 48k samples for a perfect reconstruction (assuming perfect converters) then sampling at 96k doubles the chances of sampling error, sampling at 192k quadruples it, etc.
So if we only hear and process frequences below 20k (which by Nyquist we only need 40kHz sampling rate) it is not a good idea to sample at 768kHz because the chance of a bogus sample is 16X compared to sampling at 48k (filtering out one bogus sample is not possible because any algorithm cannot determine what kind of music we are recording and processing)
The practical implication of this is: you get better results with close to perfect 48k converters than less than perfect 768k converters