Re: [ARSCLIST] Listening tests; Impact of lossy compression on audio restoration...

[Don Cox <doncox@xxxxxxxxxxxxxx>]

On 11/12/07, Jon Noring wrote:


> To comment on Howard's observation, I ran an experiment a few years
> back where I took a few WAVE files of classical and rock music, ripped
> by myself directly from audio CD's I own. I then created MP3's of each
> at various bitrates using a highly regarded MP3 compression tool
> (sorry, forgot the name of the tool), then converted the MP3's back to
> WAVE and burned the original and the MP3-sourced WAVEs back to audio
> CD for auditioning purposes.
> 
> On my lower end audiophile system (Rotel pre-amp/amp, Vanderstein
> speakers, Marantz CD-63SE player) I asked my son to randomly play the
> tracks, a sort of blind test comparison. It turned out I could tell
> the difference between lossless and lossy up to 192 kbit (although
> things got pretty subtle at 192 kbit) -- at 256 and 320 kbit, though,
> I could not reliably tell the difference. (I don't claim to have
> golden ears, so those here who have "golden ears", what do you say?)
> 
You might have been able to tell the difference at higher bit rates with
better speakers - indeed, I have seen it suggested that one way to tell
the quality of a loudspeaker is by testing what bit rate of MP3 is
distinguishable from the original.

It is likely that on little computer speakers, even low bitrate MP3s
sound no worse than original CDs (or studio master tapes).

> This tells me for "master" preservation purposes, one should probably
> stick with lossless compression such as FLAC or APE since:
> 
> 1) The difference in file sizes between the 256+ kbit and lossless
>   begins to narrow, and
> 
> 2) Digital storage continues to improve both capacity-wise and
>   price-wise, and
> 
> 3) lossy compression algorithms may adversely impact upon audio
>   restoration tools because it fiddles with the noise.
> 
> 
> To amplify #3, most of us here focus on transferring noisier analog
> sources (78s, old tapes, cylinders, etc.) So I always worry about
> lossy compression since it plays games with the spectral noise
> characteristics (as well as altering the signal some, too). To me,
> until demonstrated differently, the conservative position is to not
> apply any lossy compression during the transfer and
> mastering/restoration phases of handling audio files -- one never
> knows how this may render audio restoration tools.
> 
> So, among the audio restoration experts here, could lossy compression
> reduce the effectiveness of audio restoration tools?
> 
Certainly.

Regards
-- 
Don Cox
doncox@xxxxxxxxxxxxxx