You're in the Sound Science forum. We test things here! This issue is very important to me. I have a very large collection of music- tens of thousands of CDs, and even more records. I built a media server in my theater to serve up both movies and music. The server now contains nearly 100 TB of storage, and over a year and a half of music.
Before I embarked on digitizing my collection, I decided to test three major codecs... Fraunhofer MP3, LAME MP3 and AAC. I researched [/SIZE]artifacting. There used to be a very good page on the web with examples dedicated to cataloguing the various ways compression artifacts could occur and what the artifacts sounded like. I pulled dozens of titles from my collection from various genres, from classical to century old acoustic recordings to jazz, acoustic music, electronic music, examples from the 20s, 30s, 40s, 50s... all the way up to the most current recordings in my collection. I included digitized Sheffield Lab direct to disc LPs and downsampled high data rate audio. And I digitized each title at 128, 192, 256 and 320 CBR using each one of the three codecs.
I went over the samples with a fine tooth comb comparing it head to head with lossless, searching for artifacts. If I found one on any music sample, I eliminated that codec and data rate from further consideration. When I was done, I found that Fraunhofer 320 was *almost* transparent, but I had one CD that could make it artifact slightly. LAME MP3 was perfectly transparent at 320. And AAC was transparent at 256. I chose AAC 256 VBR as my standard setting for my music server. It took nearly two weeks full time, but it was worth it to me because I didn't want to have to ever go back and re-rip or re-encode any of my collection. I now have a massive digital library and I listen to it all the time. I have never encountered any artifact on any of the hundreds of thousands of files.
In addition to my own collection, I'm a digital archivist in charge of a large media library that belongs to a non-profit educational organization. I know a little bit about this stuff. If I was maintaining a master file- the original recording master- I would keep it in whatever format that it was created in. That way if the file needed to be remixed or edited, it would be completely flexible. But for reference files- or more specifically, files of music I listen to in my living room- compressed audio is perfectly fine as long as it's audibly transparent.
The definition of "audibly transparent" is that a copy sounds *exactly* like the original. There's no such thing as "better sounding" than transparent. There's only larger file sizes. Bigger file sizes may have a purpose if you are working in a recording studio, but for the purposes of listening to music in the home, compressed audio that meets the threshold of transparency is as good as you can get.
You can't judge sound quality by the data rate. Every codec has a different point of transparency. Some are more efficient than others.