I am doing compression comparisons between MPEG-4 AAC, MP3, MP3 Pro, RA, LQT, WMA and MusePack compression formats. I was planning on doing the comparisons with mono audio files. Are there any reasons why I should choose stereo files instead? Also, I will only be comparing differences in data reduction and frequency response. Is there something else I should be comparing? Cheers
Comparison of freq response is quite likely not very meaningful as psycho acoustic compressors typically save bits by cutting out inaudible high freq tones. This means that freq response will not be linear, while the sound may audibly just like the original. You should do your tests with stereo files, as joint-stereo algorithms are an important part of the compression process. I think you might want to study www.hydrogenaudio.org discussions and ABX listening comparison methods..
- Most music people encode is from audio CDs = stereo. - One important factor for quality/size ratio is the ability of a encoder to handle redundancy between (not totally identical) channels If you choose encoder settings properly you'll be able to encode files with each encoder that have (almost) identical sizes. Now you can compare quality. Doing it the other way round is impossible IMO. This doesn't tell you what your ears will like better. You don't compare lossy image compression's quality by converting the image to sound, do you? - Use different settings as different encoders are made with different goals (e.g. Musepack: transparency; MP3Pro: tolerable sound at ~64kbps bitrates), use VBR if possible: -- goal transparency: musepack standard, lame --alt-preset standard (extreme/insane), AAC standard + higher, Vorbis (GT3) q 5 + higher, WMA (V. 9) ??? -- ~128kbps bitrate: try what settings you need for each encoder -- ~64kbps bitrate: - " - - Double blind listening tests (ABX) are the only valid way to compare lossy audio compression. - To train your hearing you should listen to and try to ABX several known problem samples before you start your test - Go to http://www.hydrogenaudio.org/ forum - you'll find lots of valuable information about what I wrote so far there (among others in the FAQ).
The music chosen is important too. Some test samples (hydrogenaudio FAQ : http://www.hydrogenaudio.org/index.php?act=ST&f=1&t=4917 ) have been found that eventually give bad results when compressed, even with very high quality settings with which even audiophile recordings sound perfect.
