[Top] [All Lists]

Re: Digital Distortion

Subject: Re: Digital Distortion
From: Walter Knapp <>
Date: Mon, 30 Jun 2003 13:21:38 -0400
Clifford Caruthers wrote:
> Evert,
> I must have missed your post regarding this (has this been going on for
> a month?!).   If I'm not mistaken digital rounding is exactly the bad
> thing I was describing could happen, i.e. the quantization of low
> bit-rate information?  This may be a semantic difference, but I don't
> consider that damage -- you're simply mapping what little bit
> information you have to a 16bit depth, making the low resolution more
> apparent.  Is there something else going on here that I'm missing?

A lot of it is a misunderstanding of the meaning of bit depth. At 16bit,
you are encoding a dynamic range that greatly exceeds both what you find
in the outdoors, and what most folks can distinguish. Shifting the sound
within that range has fairly minimal effects. Shifting it so you
accidentally put some of it outside the limits does have major effects.

> Incidentally, doesn't this discussion support a move to  24bit depth? =

> I'm very skeptical of the need to go above 48k, but I've noticed
> greatly improved resolution at low volume with 24bit depth.  It I'm not
> mistaken, this also greatly reduces the effect of digital rounding,
> which can happen at any stage of processing...

For the actual field recording I don't see much value in greater bit
depth. With 16 bit you have some 65000 increments of sound intensity
available. I know back when I spent several years doing hearing testing
that the average person does not discriminate sound levels near that
precise. And recording practice is to adjust recorders to use pretty
much a standard level, regardless of ambient sound levels. The dynamic
range of a 16 bit recorder nearly always exceeds the range of the
environment, or the rest of the recording chain. Even after we allow a
10 - 20 dB safety pad.

However, when you get into processing the sound, computers don't carry
very many significant digits through their calculations, and sound
calculations can be a whole pile of iterations. So having software that
converts the 16 bit field data into 24 bit (or even more) for the actual
calculations will help to keep the data accurate. Once the calculations
are done this can be brought down to 16 bit for a CD without a lot of

I think this is another area where people need to be more precise in
their thinking. I see arguments for higher bit rate recorders that use
justifications that really only apply to the computer processing later
on. Recording is a different step, with different criteria. No matter
how you slice it, higher bit depth is going to fill whatever your media
is faster. Unless you use trickery, like minidisc, which uses bit
offsets to achieve most of the advantages of 24 bit in a 16 bit system
storage space.



<Prev in Thread] Current Thread [Next in Thread>

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU