Phil Leigh;613554 Wrote: 
> Except it doesn't work like that in practice because we don't record at
> 16 bits or less, we record at 24 (OK let's call it an effective 21 in
> reality) and dither down to 16, effectively masking (decorrelating) the
> quantization distortion.
> 
> Thus the effect of quantization error on low-level signals is nowhere
> near as bad as it would be if we were still (1980's) recording at 16 or
> even 14 bit!
> 
> 
> so yes, the 96dB on playback is achievable in theory - even if in
> practice it is impossible to properly record anything much quieter than
> (say)  -70dB anyway because of ambient and electronic noise ... :-)

I have a problem claiming decorrelating when working on random (music)
signals.  I understand the principle when talking about sinusoidal
signals that repeat for some period of time.  But I can't see how
decorrelation benefits a one time transient signal.  Quantization error
prevents you from perfectly reconstructing such a signal with any more
precision than the quantization step size.  Maybe a mental block on my
part...

Terry


-- 
TerryS
------------------------------------------------------------------------
TerryS's Profile: http://forums.slimdevices.com/member.php?userid=40835
View this thread: http://forums.slimdevices.com/showthread.php?t=85879

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

Reply via email to