adamdea wrote: > Arny I don't think the delta sigma thing is relevant if we are > discussing 16 bit quantisation. >
Please explain why I should believe this, as opposed to standard auio engineering texts, formal classes and lab tests. > > The ADCs will produce a single or multibit stream at several megaherz. > That is true for the internal operation of many Sigma Delta ADCs. However, up until now you've been talking about older designs. Which do you think you are talking about now, and what's our clue that you changed topics? > > Getting it to pcm is a different story. Leaving aside gettign it to 24 > bits of pcm, once it is in 24 bit form (at which it will probably be > mastered) it WILL be downsampled in software to 16 bits. > Tht is true only if the distribution medium handles 24 bits. For example, that excludes DSD but you don't seem to be able to qualify what you say for the limited set circumstances that it applies to. > > Delta sigma does not come into it. > I thought you were talking about ADCs, not mastering software. When did you change that? If you are talking about ADCs then how Delta Sigma works is very relevant because its how quantization error and dither are handled. > > The downsampling may or may not be done using noise shaping > That may be true of some mastering software, but most mastering software emphasizes their options for noise shaping. Noise shaping during mastering provides a means to produce the subjective equivalent of up to 120 dB dynamic range with 16 bit recordings. Most well informed persons who are engaged in mastering are interested in advantages of that nature. If you critize 16 bit media while intentionally excluding well-known effective technololgy such as noise shaping then you have seem to have the makings of a straw man arugment. > > but is quantisation in the sense described in the text. > The text is incomplete and does not mentioin more recent but generally accepted technology. I hope that its authors are doing that to mislead people, but rather are doing it to keep the story as simple as possible for newbies. > > The point remains that there is nothing particularly special about > adding quantisation noise of the same power as the existing noise in the > channel. The above still ignores the fact that quantization noise is not intentially added, but comes into the picture as an inherent characteristic of digitization. Using your own reference: http://www.dspguide.com/ch3/1.htm "When isn't this model of quantization valid? Only when the quantization error cannot be treated as random." I.ve shown that most modern quality ADCs (e.g. Sigma Delta) for audio do indeed treat quantization error as random. It appears that you are cherry picking information from your reference to bolster some erroneous thinking, and apparently ignoring parts of your own reference that don't agree with your erroneous ideas about quantization error. ------------------------------------------------------------------------ arnyk's Profile: http://forums.slimdevices.com/member.php?userid=64365 View this thread: http://forums.slimdevices.com/showthread.php?t=105717 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/audiophiles