opaqueice;289120 Wrote: 
> I already did: you didn't enjoy your breakfast that morning.
> 
> It's very, very simple - if you take a 24 bit binary number and divide
> it by some factor, the error you make will be in the "25th" bit (if you
> round properly).  If you truncate rather than rounding, the error may be
> in the 24th bit, but that's as bad as it can possibly be.  So the error
> - I'm talking about the absolute error, *NOT* the error as a percentage
> of the signal - is never larger than 000000000000000000000001.
> 
> So all we have to ask is whether 000000000000000000000001 is ever
> audible on a stereo system.  It has the best chance with the analogue
> gain maxed and with a frequency in the good part of the human hearing
> range, but as I've shown it remains waaaaay below the threshold of
> audibility even then.  We might also want to ask whether (say)
> 010000110000000000000001 is audibly different from
> 010000110000000000000001 + 000000000000000000000001.  It's not - it's
> -harder- to hear small changes in level than it is to hear the change
> alone (because of masking), and in any case my argument proves that it
> would be *physically impossible* to hear that change either in
> isolation OR added to another signal.
> 
> You are probably confusing S/N - which will be much lower than 144dB in
> these cases, since the original data was 16 bit - with audibility of the
> distortion caused by rounding.  Let me try to clarify that for you.  
> 
> Here are two ways in which distortion can be *inaudible*:
> 
> 1) The S/N is sufficiently high that you do not hear the
> noise/distortion no matter how loud the volume is, because the signal
> always masks the distortion.  That is one case where we don't have to
> worry about distortion.  A good example of that is the noise floor of a
> decent digital audio system while playing music.
> 
> 2) The maximum possible level of the noise/distortion is too low to be
> audible *even when it's played without the signal*!  In this case it
> will *never* be audible - even when the S/N ratio is *0*.  There are
> *no exceptions*, and that's what I've shown is the case of a rounding
> error in 24 bit audio.  Of course if you had -infinite analogue gain-
> available to you, you could always turn it up to the point that it
> would be audible - but you don't.

You are still thinking in your own paradigm.  I'm sure they must have
taught you at Professor-school, to always question your own
assumptions?


-- 
Patrick Dixon

www.at-tunes.co.uk
------------------------------------------------------------------------
Patrick Dixon's Profile: http://forums.slimdevices.com/member.php?userid=90
View this thread: http://forums.slimdevices.com/showthread.php?t=45736

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/lists/listinfo/audiophiles

Reply via email to