View Single Post
Old 6th December 2013
Originally Posted by jalcide View Post
mp3 "compression" is lossy data compression, not audio compression (that modifies signal gain/volume).

the only thing the two have in common are letters, and the tendency to make a perfectly good mix sound like crap.
Thank you... I saw more than one reference to mp3 compression.

To further the discussion of data compression...

Data compression means that when an audio file is created, it is turned into a bunch of 1s and 0s. WAVs and AIFFs are not compressed at all. Most mp3s are lossy compression. Then there are lossless compression files.

Lossy compression means that the file is scanned, and any repetitive information is deleted. So, lets say for every snare hit (this is oversimplified, but its an example that makes the point), every bit of information that represents every snare hit is scanned, and anytime a series of 1s and 0s is repeated in every single snare hit throughout a song, it is kept for the first instance and for flagged. It is then removed for every other snare hit when the file is compressed (made smaller). When you play it back, the mp3 player then reads the first snare hit and the flag, and then reconstructs the audio file in real time putting that deleted information back as the song is played. So, on one hand, there is no real change because as long as the 1s and 0s are put back like they were, there is literally NO difference in the file before or after.

The problem is... music files are very complicated. It has to read (again just an example) every single little flag for every single sound and put it all back as the song plays for every sound in real time. If you we talking a series of numbers this long... 100110100101001010001000101010 it wouldn't really be a problem. The reality is, its a series of numbers much bigger than that stacked all on top of each other for every single milisecond of delay repeats, reverb wash, every frequency present, every volume change of every frequency present, etc. etc. etc. In reality, an mp3 player just can't do this in real time, can't even really pre unpack the file fast enough in relation to songs being a few minutes long. Errors are made.

As mentioned, this is heard is some pre-transient accuracy loss, and is also heard in the areas where the most amount of file deletion and restoration occur... low frequency and high frequency. The lows seem a little more flabby because they're more often being "put back in" than mid frequencies are. Highs seem less crisp for the same reason. Transients are affected in the way described because they are, by their nature, in the proximity effect low end land, and snappy high end land.

In this thread's context, it would affect the results more to hear an mp3 as the source that is compressed, than it would to hear a compressed WAV converted to mp3.

Most people, and this is why mp3s are not being thrown out the window en-masse, simply do not have the equipment to hear many of these things, nor are they really picky enough to notice them if they did. If something doesn't make an obvious flutter, or click, or pop, most people won't really notice it....especially since there is now nearly a whole generation that has never really even heard a WAV file on a cd to have a frame of reference. Also, with modern mastering methods, its not like you're going to play someone a cd and there be the kind of difference for most people to hear the way there would be if you played them a great Hi Def file or a vinyl of the same song where you can easily point out more exciting and up front drums, silkier highs, and less farty lows.. not to mention the excitement of having a mix grow as the music grows.

Not surprisingly, you'll tend to hate an mp3 of an original master cd of something like say Hysteria's original cd release than you will any of its re releases, or anything modern in the brickwall mastering era. It is both more work to compress that audio file, because it is more dynamic with more changes, and more noticeable because of the lack of transient squashing done in older mastering. If you've ever wondered why artists are now "mastering for Itunes", now you can start to imagine why. You also now can imagine, in a tangible way, why people like us are starting to be rather vocal about for Itunes mastering being better than for CD mastering. They're purposefully mastering a version that will be less affected in those areas where most of the file deletion occurs. Really, its just like mastering used to be... when mastering was done for records, they had to master things. If it was too loud, the grooves would blow out. They had to take out the bass, and the exact bass frequencies that the turntable's filter would put back in, so that the grooves wouldn't blow out. You simply could not send a signal from the 2 bus straight to a lathe. In the CD age.. aside from making things as loud as possible, mastering is virtually unnecessary for some high end mixes. I know I will be slammed for this, but I'd bet money that the unmastered mixes of my bands songs would be indistinguishable to many a super experienced ear from the mastered versions. The only reason is the mix and mastering was done in a prepared manner to get the modern loudness with as much old school dynamics as possible. Compromises made yes, but mastering isn't going to fix a bad and unpunchy mix anymore than mixing will fix a bad and unpunchy recording. When you hear masters that have had to compensate for flat mixes, you can usually tell for this very reason. When you think of that magic commercial frequency curve you see on an analyzer, all the great mixers in the world send the mix to be mastered with that curve 95% or more there. Its just a day where more people are more aware of more information.

Please note that I did say "VIRTUALLY" unnecessary. LOL

I made note of the subsequent tests, and I need to give them a listen because I couldn't even come to a conclusion off of the first test. #2 felt pumpy and squashed, but it seemed snappy too... odd how those things work. It started making sense when I re read that the settings were different. Fabrice's post is also invaluable, even if a whole lot of science to absorb. LOL