Rene Herman wrote: > André Pouliot wrote: > >> Sorry to tell you but using 16 bit 44,1KHz sampling most person can >> tell the difference you are already at twice the nyquist frequency >> for what the ear can listen. > > > No, the Nyquist frequency is the highest frequency that can be fully > reconstructed from data sampled at a given rate, and is half that > rate. With a rate of 44100 therefore, 22050 is that highest frequency > and this is not very far above the the highest frequency a good, > young, human ear can hear. 44100 is about minimum for good quality, > full spectrum audio therefore. > > In any case, sampling at 96 or 192K is not done to now be able to > encode 48 or 96K signals, but due to the preconditions of Nyquist's > theorem. Nyquist only holds (the signal is only losslesly > reconstructable) when the signal is very strictly limited to half the > sampling frequency. In the analog domain though, there is no such > thing as an infinitely steep filter meaning you still can get aliasing > efects from higher frequencies that were present in the pre-adc signal. > > The higher the sampling rate, the easier for a filter to filter out > frequencies that would alias into the audible band therefore. 96/192 > is not audiophile humbug or anything. > > Rene. > Sorry I didn't make myself clear I tend to drop part of what I think and not write it. I know nyquist theorem, what I want to say was more like for no aliasing effect in normal use you more than make the requirement. Since a normal human ear could ear theorically up to 20KHz it's more frequent to encounter people who ear up to aproximatly 16Khz. Also since your at the upper bound of the range event if you don't a have a well define signal with many sampling point, people would most likely will not ear a difference. But if you input a higher frequency that the anti-alliasing filter didn't clean out and have enough energy to still be detected, sure you get a different frequency reproduced at the output. In that case you need a higher frequency sampling rate to detect it and filter out the higher frequency signal.
But normaly the anti-aliasing filter is made so you have minimally -3dB at the cut-off frequency, the farther away you get the more the signal is lowered. So at 22,05Khz you start to get a reflexion of the signal. If we assume that someone ear a reflected signal at 20Khz you must enter a signal at 24,1Khz so he start to ear something, at that point the signal probably as an attenuation of at least 6dB. So if someone ear the deformed signal at that frequency most likely it will go as an harmonic of the main signal since most of the energy in audio signal is at a lower frequency. If you make recording for a studio it do make sense to use a higher sampling rate so the higher harmonic (exemple : a guitar) will not go on the recording and be cancelled, it's useless information(except for making a dog flee). But that's mostly a moot point except in a number war for home user. Most people in Linux mostly play music or game, a card with well isolated input and output from noise you don't need anything else than 44,1Khz sampling and 16 bit precision. If you try making a recording studio equipement in that case going for 192Khz 24bit in an outside box could be good before reworking the signal. But once you have a clean signal even if you downsample it at 44,1Khz 16bit, with a first order output filter you'll get good enough result. No one could really tell the difference except the maybe 0,1% of the population who have exceptionnal earing. If someone try to record is voice with a sound card you have more quality than necessary since the human voice is under the 1Khz. It's all depend of your target for a sound card. As most other post in the mailing list seem to point it's more a card for professional so the higher the better but going more than 24bit 192Khz is not useful. If you want to make a card for the linux market most people don't need more than the classic 44,1Khz 16 bit(with a good protection from noise) so most old card do the job. But it's sure would be nice to use a card who have realtime dolby digital encoding, an optical connector output and work on any platform. André _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
