[Flac] Low compression on high noise?

Brian Willoughby brianw at sounds.wa.com
Sat Feb 19 17:26:04 PST 2005


Hi Thomas,

The amplitude of the piece has the most to do with the ability of FLAC to  
compress it.  Second to amplitude would be frequency, i.e. higher frequency  
content would compress less than lower frequency material.

FLAC starts by converting the absolute samples to differential samples.  Lower  
amplitude signal have smaller differences between each sample, and thus take  
less space.  Also, lower frequency signals change very slowly, making the  
difference between each sample smaller.  Lots of high frequencies, such as  
noise, would tend to result in large changes between samples.

I do a lot of live recording, and usually the recording level is set low  
enough to ensure that none of the samples clip.  These raw, unprocessed  
recordings compress very well with FLAC.  However, mastering for CD usually  
involves boosting the average level, followed by compression and/or limited to  
kill peaks that go over 0 dBFS.  The mastered recordings usually produce larger  
FLAC files, even thought they are "compressed."  This is because the amplitude  
has been raised.

I bet that if you look at your Merzbow and Japanese noise in a wave editor,  
you'll see almost solid black samples!  I assume this music is designed to be  
played loud, and the material is high amplitude and high frequency.

Brian Willoughby
Sound Consulting


Begin forwarded message:

Hi List,
I noted that compressing music like Merzbow or other Japanese chaotic
noise-bands usually results in very low compression rates.
That means, that the resulting .flac-file is nearly as big as the
initial .wav-file.
Other kinds of music get compressed much better.
Any explanation?


More information about the Flac mailing list