Hi,

I was just wondering why there is a bitrate like 320 kbps.

Personally I don't hear the difference between 192 and 320 kbps, even in "critical" music like Yes' 70's albums. Or does it depend on other factors?

Roughly speaking HA members fall into 2 categories: practically minded, and a bit paranoid minded (like me).

For the practically minded everything is fine in the say 128 to 192 kbps or so range (and most of the time this is the case also for the paranoid minded).

The paranoid minded care about nontranparency in bad situations. For very serious pre-echo problems they have to accept that the result won't be transparent (though there are differences with encoders and versions), but other than that they can expect of a good encoder that 320 kbps results are transparent.

Two examples:
harpsichord music is a problem for many codecs, and especially for mp3. I don't know any mp3 encoder that gets a sample like harp40_1 transparent at 192 kbps.
Moreover we have the phenomenon that with encoder development it can happen that there is significant progress on one hand, but also serious regress in specific aspects. Worse so, the regression may not be noticed for some time. We had this with 3.97 which came out great in a 128 kbps listening test but has tonal problems (like the 'sandpaper noise' problem) which are considered serious by not only paranoid minded.
With 3.98 development these problems are overcome, but imagine based on the great results 3.97 showed for a long period you encoded your collection using 3.97 -V2, you are left with this problem. No problem if you have archived everything losslessly, but someone like me who has a lossless backup only for the very good tracks of his collection, has a problem with his archive.
However if you decide to use 320 kbps (or a similar extremely high quality setting) there is a large headroom to catch up encoder problems. In fact 3.97 CBR 320 may not be transparent with the specific problems (to me it is), but it's hard to imagine that the quality is not fully acceptable to anybody.

So there's actually no need for > 192 kbps if you aren't paranoid, encode everything with 3.98 or don't have a need for harpsichord music?

It ain't that easy. harpsichord was just a sample. Imagine a problem will show up later with 3.98 the way the tonal problems showed up with 3.97.
In fact there is a problem left with current Lame development (scheduled to be targeted with the next version): something I call a 'tremolo problem'. It happens for instance with the sample 'trumpet_myPrince' or - worse - lead-voice. IMO it's acceptable with -V2 but it's not transparent. The problem is not introduced with 3.98 but existed before. It's also a more general problem and is not Lame-specific, but it's more pronounced with Lame than for instance with FhG encoders.
And if you consider using an FhG encoder, you'll probably run into the the same situation that earlier or later you'll find a sample where FhG behaves a bit unadequately compared to other encoders. eig is such a sample.

In the end it comes to the same thing: if you are practically minded there's no reason not to be content with 3.98 -V2, and for the very cautious it still makes sense to go beyond.

You also have to factor in "usability". High bitrate encoding removes less of the original signal than low bitrate encoding, so it more closely represents the original signal. A high bitrate file can better "tolerate" additional encoding passes, which is handy if you ever have the need to transcode (say, if you need to go from MP3 to AAC, or vice versa, or from MP3 to OGG, etc.).

There might not be an audible difference (to you, or to others) between 128kbps and 320kbps in some cases, but the 320kbps file will be more representative of the original data. Sometimes that's important, and sometimes that isn't important at all. It depends specifically on for what reason you're encoding in the first place.

If you're going from a lossless copy to a lossy copy for your jukebox/portable/whatever, then don't be afraid to go with the lower bitrates if file size is critical, and if transparency is less critical. Keep in mind that just because you may be able to A/B the original file from the compressed file doesn't mean that the compressed file is necessarily unusable. The artifacts introduced may not be annoying or off-putting.

I always encode with the aim to have a high transparency. But I haven't heard any differences so far...

Obviously you belong to the practically minded people. So everything is fine using 192 kbps.

I hear a difference between 192 and 320, and even more so between 320 and lossless.

Since lossless isn't practical (for an ipod or your hard drive), I go with 320 cd rips for all my music.

You cannot logically hear more difference between 320kbps CBR MP3 and lossless.

In order to adhere to out terms of service you need to provide ABX records to back up such statements, and not violate TOS #8.