[vorbis-dev] Spectral band replication

[Gregory Maxwell]

On Mon, May 14, 2001 at 06:45:23PM +0200, Robert Voigt wrote:
> On Monday 14 May 2001 13:51, Jan.Tangring at et.se wrote:
> > This seems to contradict, and with a margin, the current statement of
> > Thomson multimedia. Thomson multimedia says MP3PRO 64 kbps sounds like MP3
> > at 128 kbps. Thomson is marketing MP3PRO as a technology for doubling the
> > capacity of portable MP-equipment.
> 
> I'll tell you how I would obtain to figures like this if I were a greedy 
> inventor wanting to make as much money as possible. I would select some 
> pieces of music where my new audio codec performs very well and do double 
> blind tests with low quality equipment, and claim that this is the kind of 
> equipment most people use for listening to compressed music (computer 
> speakers, headphones that come bundled with mp3 players). Of course my test 

Actually, although not scientifically tested (just my experience), perceptual
codecs work *BETTER* with high-end equipment. The listening tests that form
their basics are done on good equipment, and they compress based on what they
think the listeners ears will be receiving (masking, etc). 

For an example of this, play a .wav and a .mp3 in XMMS and then make the EQ
settings funky (simulating bad equipment), listen to the difference of the
difference. A transparent mp3 is no longer transparent on the funkey EQ
settings because you have thrown off the masking calculations, MP3s subband
coarseness makes this more obvious, but the effect should exist for all
perceptual codecs..

Unfortunately, since the precise nature of the bad as it varies from place to
place, you can't test or train for it.. Your best hope is to design for a
perfect sound system and hope that any users who actually care about quality
will create a good sound system.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.