192k vs. 128k:

Was 192 as high as you went with the tests?

It might be fun to take one or two pieces which exhibited the worst aliasing (to your ear) and try them at higher bitrates, or at highest-quality VBR rates using LAME or Xing.

When I encode in high-quality VBR, I've seen some frames jump to 320kbps...

Biggest difference in the bass is that it looses its firmness and less defined;

I'm starting to hear more and more people say this. It sounds illogical to me (the low frequencies should be reproduced quite well at any bit rate), but these comments are starting to reach a critical mass. I haven't ever noticed any problems with the bass, but now I'll start listening more closely.

In some cases, you can attribute those comments to the fact that many people are previewing the MP3s on different sound reproduction equipment than the CDs they're comparing them to. But in your case you did the Right Thing and tested the encoder only by decoding the MP3 back down to a wav file and burning that to a CD. So it's not an equipment thing in your case.

Same applies, but in higher order; particularly narrowing of soundstage was noticeable. Stuff that was firmly located to the left made a shift inwards.

Important information for the "soundstage" discussion we were having a while ago. Of course, this is the MP3-soundstage-limitation question, not the Empeg-soundstage-limitation question.

Also, when listening in the car imaging and soundstaging in the audiophile sense become pretty moot anyway.

Don't let Doug hear you say that! Many of us car audio enthusiasts have better sound systems in our cars than in our homes.

It effects of datareduction appear to wreak the greatest havoc with the classical piece, least with the fairly uncomplicated track with woman's voice.

Interesting. I've noticed that solo voice can really show up artifacting. Just yesterday I was noticing some compression artifacts on a solo female voice track (The album happened to be "Mouth Music" by Swan and Mackenzie). I didn't think the track was complex enough to cause audible artifacting, the voice was firmly in the midrange and very clean. But what it showed was that having all the "air" around the voice allowed faint artifacts to become audible, especially on note decays, that would otherwise have been buried in the rest of the mixed instruments.

And this was egregious artifacting, too. It wasn't subtle. You know that gurgly, alien-talking-underwater kind of sound you hear on really-heavily-compressed streaming RealAudio tracks? I could hear that as her voice trailed off on certain notes. And this is at 128k, not at lower bitrates where you would expect to hear it.

Mind you, I had to be listening at full volume to hear it. And I was running the voladj kernel, so it was cranking up the note decays well above the noise floor, so it's kind of unfair, I guess. Still, it was there.

___________
Tony Fabris