mp3 format questions

[VJ]

Hello,

I have a questions regarding audio formats... If I convert a 320kbps MP3 file to an uncompressed format (e.g. FLAC), and then convert this FLAC to a 320kbps MP3 file, how will it compare to the original? Same quality or worse?

I think I know the answer (I'd say worse), but wanted confirmation. The reason I ask is that I have my entire CD collection in FLAC format, apart from some mp3 files which I can't rip (damaged cd); quality permitting I'd like to put it all in FLAC...


Jörg

[Umfriend]

Simply try it once and do a binary comparison of the two mp3 files.

I am ignorant in this matter but I assume decoding an mp3 gives a binary data stream that then goes through a DAC. If that is the case then I would expect the resulting mp3 to be the same provided the actual resolution of the uncompressed file is higher (at all frequencies). Ignoring rounding issues, compressing it again should give the same as the original mp3 because otherwise it would mean that the original mp3 is not the best obtainable result from the source (or the decompression is not the optimal algorithm).

But then, I could be so terribly wrong.

[dZeus]

Simply try it once and do a binary comparison of the two mp3 files.

I am ignorant in this matter but I assume decoding an mp3 gives a binary data stream that then goes through a DAC. If that is the case then I would expect the resulting mp3 to be the same provided the actual resolution of the uncompressed file is higher (at all frequencies). Ignoring rounding issues, compressing it again should give the same as the original mp3 because otherwise it would mean that the original mp3 is not the best obtainable result from the source (or the decompression is not the optimal algorithm).

But then, I could be so terribly wrong.

how I understand it:
decode should give you identical results between properly written mp3 decoders.

However, re-encoding might give you a loss of information due to several factors:
- psycho-acoustic settings used might differ between the two encoders used to encode at 320bps. (they differ for encode settings within an encoder build and often at same settings between encoder builds).
- psycho-acoustic algorithm might provide with different results when run from the original file containing the information that will be partly removed, resulting in a potential loss of information in the second run on the file

[Umfriend]

I'm not sure I understand the second issue. I assume that with the initial compression from a superior format, some information is removed but that the next time it is compressed (after being decompressed in a superior format), that information is no longer present. Why would a second run require more information to be lost?

I do regard using different settings as using different algorithms.

[VJ]

dZeus second argument basically:
original -> mp3 : some information is lost/omitted
mp3 -> flac : basically same information as the mp3, but will differ from original
flac -> new mp3: source differs from original, so the compression algorithm may work differently

Esp. combined with his first argument, as I don't know with which version and settings the original files were encoded...

Guess I'll keep it in mp3.

Jörg

[dZeus]

I'm not sure I understand the second issue. I assume that with the initial compression from a superior format, some information is removed but that the next time it is compressed (after being decompressed in a superior format), that information is no longer present. Why would a second run require more information to be lost?

I do regard using different settings as using different algorithms.

well I have no exact knowledge about psycho-acoustic models used in encoders. What it comes down to is, how absolute or relative the model is.

e.g. it might regard a loss of 0.1% of the sound worthwhile if it saves some bandwidth. This loss of 0.1% is a different 0.1% when performed on the resultant file than on the original.

That, and who knows whether those mp3 encoders are as bug-free in these environments as they could be. At least not too long ago there often were problems where (video/audio) encoders could assign more bitrate to certain passages than effectively could be used. That means that above a certain bitrate, theoretical quality/resolution would not increase but was still taken up.

[VJ]

Yes, it is a complex issue (hence the question for some feedback ).
(I had a feeling it would be best to keep them in mp3, but just wanted to be sure)

My idea is to keep my library in FLAC, fully tagged and classified (and backed up ); and use uncompressed where possible.
Depending on the target where I intend to play the files (iPod, mobile phone, ...), I batch convert them to the best format (both regarding container and compression: i.e. m4a for iPod, wma for mobile phone, ...). The conversion is relatively fast (my collection converts in 7 hours on my dual Xeon, I'm guessing just over 2 hours on my Core i7) and requires no user interaction (I can really recommend dbpoweramp).

Regarding the conversion for iPod: how does the Nero AAC encoder stack up to iTunes?
(I only found some older reviews, in which iTunes had the better encoder...)

Jörg

[VJ]

Apperently, the Nero AAC encoder looses the tags, I'm stuck with using iTunes to convert my lossless apple format to AAC. The downside is that iTunes only converts a single file at a time, whereas the other software works in parallel on multi-core systems.

I have started converting files to AAC using iTunes (and expect it to finish this evening), but was wondering: is there much difference (if any) in quality, battery life and file-size between the two format (when using the same settings of course)?

Jörg