If the layer 3 implementations were done perfectly, a 256kbit layer 3 stream would sound better than a 256kbit layer 2 stream - but it's not done perfectly. It's nearly impossible to say for certain whether the ear can hear a frequency or not without massive amounts of calculation, which just isn't done because it would take forever to compress.

I doubt you could get 5:1 lossless compression, maybe 2:1 at the most. On completely random data you can't lossless compress anyway. Sound happens to be quite a random signal.

Not sure what you mean by 'compressing globally'.

- John.

(The above may not represent the views of Empeg :)