| |
|
The Creative Xmod
it's a white box with a big shiny knob on which plugs into USB sockets and "makes mp3s sound better than CDs". apparently
and there's this graph:
I'm guessing the axes, they've forgotten to label are frequency and that well known empirical measurement 'betterness'.
Assuming that it can make mp3s sound "better" than CDs, then what's it doing?
As the mp3 algorithm is based on psychoacoustic masking effects, it removes information within audio range, then 'better' can't mean more accurate- unless it accesses a database which has all the remaining audio information at a greater resolution to CDs?
As it doesn't say it needs an internet connection or a hard drive large enough to contain all the missing audio information for every piece of audio ever recorded by humankind (and yet to be recorded).
So, it must do something else to 'betterise' mp3s:
"It intelligently enhances the highs and lows, restoring detail and vibrance to your audio."
i.e. The algorithm makes things up
now, what does it makes up?
What I first thought was to frequency analyse the least compressed bands and identify patterns in frequency within these bands and continue them into more compressed bands (and possibly filtering them), which would technically make it a synthesiser based on pre existing sounds (interesting idea, may have to follow that up). Unless this has got a pretty meaty processor I doubt it's going to be doing any FFT analysis in realtime, so that's a no.
Instead, what I think it just does is to add noise to the higher frequency bands when there's some activity there and possibly add dynamics processing to to the lower bands, making the bass louder and any filtering any other noises/ artefacts/detail.
Or, if the designers have been hanging around valve amp enthusiasts, add some 2nd order harmonic distortion.
Or maybe it's the big knob so that "Dynamics are better and overall naturalness is improved."
I'm yet to listen to many of the examples yet, as the flash demo thing takes ages and it's on dodgy laptop speakers, let alone put samples through any kind of analysis (would involve feeding the soundcard output back into something else).
Also, this is probably great for listening to certain records, but I dread to think what it'll do to a non-compressed classical recording (have you ever accidentally 'normalised' one of those?)
any other ideas? |
|
|
