bluegaspode;694728 Wrote: > No - quite the opposite. I do believe that Nyquist is right and that in > theory all this information is enough to recreate the original signal. > > > But for some reason I don't believe (ok - lets say I'm at least > suspicious) if DACs are able to fully adhere to the Nyquist theory and > are capable of recreating the signal exactly. > > So why is it considered to be better to have > a) an oversampling DAC that (due to the Nyquist theorem) can > approximate the missing information quite correctly (but due to limited > processing power it will only be approximated) > > b) a DAC that is fed with 192kHz from the very beginning, so wouldn't > need an approximated oversampling > > Or put it differently: > Why is a DAC that is given the string "the cat is black" and > oversamples it to "tttthheee ccccaattt iiiiss bbbbllaaaccckkk" (please > note the approximation errors in each word that I introduced with my > oversampling algorithm) considered to produce better/the same results > as a DAC that is given the more correct "ttthhheee cccaaattt iiisss > bbblllaaaccckkk" as input from the very beginning? > > Why are people thinking that I don't believe in the Nyquist theorem? > The theorem has nothing to do with the situation that DACs (for > whatever reason) are over/upsampling nowadays internally. > And if that is needed than using the correct input with a higher sample > rate might yield better results than DACs introducing approximated > oversampled values? > > >
There are a number of points mixed up here. 1 The real point about the articel cited in the OP is that it argues that if your downstream equipment is only specified at 20KHz, trying to reproduce anything above 20Khz is dangerous - you're just asking it to create spuria above nyquist - but due to intermodulation distortion is will INEVITABLY create spuria in the audible range 2 It is thereofre dangerous NOT to filter out everything above the spec of the downstream equipment. 3 now once you are doing that, what is the point of hi rez file? You are incurring huge data redundancy as compared with the 16/44 which you can upsample/oversample to get the benefit of the same digital filters which you would use with the high res. Remember you are only approximating the data you are going to filter out above nyquist. It is true that the higher sample rate material has a slightly lower noise level in the audible spectrum, but 16 bit is quite good enough and 24 bits is more than enough already. Using high sample rates is a very inefficient way of lowering the noise floor in the audible range for the storage/transmission medium (it is a different matter at conversion- hence delta sigma dacs) 17/44 has the same noise floor in the audible range as 16/88 but a much lower data rate. 4 also don't forget that we can continue to get the benefit of upsampling well beyond 192k. Childsplay nowadays- but you would be having insanely huge files if you insisted on using those high sample rates natively in the stoage transmission mechanism. -- adamdea ------------------------------------------------------------------------ adamdea's Profile: http://forums.slimdevices.com/member.php?userid=37603 View this thread: http://forums.slimdevices.com/showthread.php?t=93990 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
