PDA

View Full Version : Compression Formats



Ducatista47
05-14-2006, 01:03 PM
I just want to suggest that there are much better formats for compressing audio than mp3, and I would like to recommed one, FLAC. It is a marked improvement to my ears.
http://flac.sourceforge.net/

Clark in Peoria

Hoerninger
05-14-2006, 01:45 PM
Thank you for the tipp, it's pleasant that there is even a port to OS/2 besides many other OS.
" FLAC stands for Free Lossless Audio Codec. Grossly oversimplified, FLAC is similar to MP3, but lossless"
MP3 has losses even in high quality modes. The longer you listen the better you will hear it. So I don't use (and need) it anymore.
___________
Regards
Peter

lfh
05-14-2006, 02:54 PM
FLAC and MP3 are very different animals: The former is a lossless codec, i.e. the output after decoding is bitidentical to the input file (compare e.g. to the zip format), whereas the latter is a lossy codec that relies on psycho-acoustics to achieve a high degree of compression. Unfortunately lossless codecs can only reduce the file size to about half the size of the original.

It's important to note that the "MP3" standard (MPEG-1/2 Layer 3) only specifies the bitstream format and the decoding algorithm - it's up to the implementer to design the encoder. The idea behind this approach is to allow for encoder improvements such that e.g. more sophisticated psycho-acoustic models can be incorporated and the available bits be put to better use within the given constraints. Hence the audio quality of a given tune at a given bitrate varies significantly from encoder to encoder.

MP3 is a rather dated format. If a higher compression than that of the lossless codecs is called for, the successor to MP3, i.e. MPEG-4 AAC (Advanced Audio Coding) is a very good choice. MP3 at 128 kbit/s has audible artifacts even using the best encoders, but AAC at 128 kbit/s (again assuming a good encoder) can be very hard to tell apart from the original - even for trained ears and when using a high quality playback system.

That said - for reference storage I'd also go with a lossless format such as FLAC, but for everyday listening (e.g. using the iPod) AAC in the 128 - 160 kbit/s range is just fine and far superior to MP3 at the same bitrate.

Regards,
Fredrik

Ducatista47
05-14-2006, 07:53 PM
There is one other good use for FLAC that I know of, Fredrik. I discovered it through file sharing. If you put up a FLAC instead of an mp3 or AAC, whoever receives the file has choices of what to do with it not available to other less fortunate sharers.

Thanks for the tip on AAC!

Clark

louped garouv
05-16-2006, 04:47 PM
I tend to like the WAV files that i get from some friends that remix or produce electronic music professionally....


here is a tid bit that I posted a few months back regarding a test of formats done at London nightclub Fabric...
http://www.audioheritage.org/vbulletin/showpost.php?p=89835&postcount=8

lfh
05-16-2006, 05:57 PM
I tend to like the WAV files that i get from some friends that remix or produce electronic music professionally....


here is a tid bit that I posted a few months back regarding a test of formats done at London nightclub Fabric...
http://www.audioheritage.org/vbulletin/showpost.php?p=89835&postcount=8

WAV is normally (but no necessarily - other formats are possible, too) plain 16 bit PCM, just like on CDs.

As to informal tests such as the one linked to I'd say most of them are completely useless. In order for a test to be meaningful, it must be (double) blind, the coded items carefully checked wrt potential level differences and compensated for if needed, the number of listeners sufficiently large, the instructions on how to grade the codecs under test clear (and understood), preferably contain hidden references such that "deaf" listeners can be excluded, test conditions and specific encoders used well documented, confidence intervals calculated for the scores, :blah:

Here are links to hobbyist tests that appear to be at least somewhat reliable:

http://wiki.hydrogenaudio.org/index.php?title=Listening_Tests

http://wiki.hydrogenaudio.org/index.php?title=Lossless_comparison

Although dated, here's the report from the EBU test conducted back in 2000, which also describes the MUSHRA test commonly used in professional tests

http://www.ebu.ch/en/technical/trev/trev_283-kozamernik.pdf

In 2003 EBU tested codecs at lower bitrates, see

http://www.ebu.ch/CMSimages/en/tec_doc_t3296_tcm6-10497.pdf