Audio spectrum analysis

[Keiyakusha]

Hi! I'd like to ask your opinion about spectrum on this images: 1 (http://img200.imageshack.us/img200/8472/72085352.png), 2 (http://img185.imageshack.us/img185/9417/91839061.png).
Which one looks better? In my opinion, the 2nd one is better because lower but "clearer" frequency is better then higher but more thinned(lowpassed?). What do you think?

[Slogra]

The second seem to have more high frequencies... but that doesn't say much about the actual sound quality. Try listening to it instead.

[Keiyakusha]

If only I could hear the difference here. ^_^
Maybe I should rephrase myself a bit. These images are just example of what I want to know: what is better theoretically to keep higher but thinned frequencies or lower but solid ones? I think the second, because generally for human ear it's easier to hear lower frequencies.

[bratao]

I think that the second is too artificial.
The second is much more closer to what the natural soundwave should be.

Why you don´t try to compare with the spectrum of the original ?

[Keiyakusha]

I think that the second is too artificial.
The second is much more closer to what the natural soundwave should be.
If I understand you correctly: second = too artificial, second = good, too artificial = good?
Why you don´t try to compare with the spectrum of the original ?

Because my question not about comparing 2 tracks, but about encoding in general... These 2 image in 1st post have only one purpose - to illustrate different ways of lossy encoding.

[twist3d]

"Ultimately, however, no visual representation is going to do an adequate job of showing what has really been lost in the lossy encoding."
I suggest you to read some hydrogenaudio forums (http://www.hydrogenaudio.org/forums/) for ABX blind listening tests vs. analyzing spectrum graphics.
As a rule of the thumb however, you can't judge sound quality out of spectrum graphics.

[2Bdecided]

...the biggest potential problems with any lossy codec, except for the low pass frequency, are going to be hidden on spectral plots like these.

You could add bucket loads of easily audible noise within the most audible range, and it would be invisible on these plots. Dramatic temporal smearing of transients would be invisible too.

Definitely read the hydrogen audio FAQ.

Cheers,
David.