a question on MP3 conversion

[Toki987]

I`m sure there is a very scientific and correct technical answer somewhere in a book or on the web that details the conversion process, which most likely is too technical and filled with acronymns to the point it would choke most. I know that there is dithering thats involved to compress the amount of data representing all the aspects of the recorded information in order that the file size is smaller. The part I`m not at all sure or familiar with is, What other manipulations are done during the process?
After watching many of my wav`s converted to MP3 and listening later, it sounds as if some sort of standard algorythm is used as far as "peak level, compression, limiting, and eq". I wondering if someone could explain it in somewhat simpler craftsman terms, as to what is actually going on in the process.

 

[Hawking]

Without getting too technical, the basis for the MP3 format uses something call perceptual modeling. Basically, it knows the frequency response of the human ear and EQ's the signal to match it. Then it compresses the file (not audio compression) based on what information isn't absolutly necessary to recover the data in usable form. The key to the whole thing is that it's not the same each time. If you take 2 WAV files of different music but the same exact file size, and convert them both to MP3, the resulting files will not be the same size any more. The MP3 technique changes based on the content of the file being compressed. That's what makes MP3 so good. It changes slightly to adjust for different signals to obtain the best results every time.

In short, you are correct. There's EQ, compression (both audio and data) and limiting happening in different amounts.

Here's some further reading that gets a little more technical if you're interested.

http://www.iis.fhg.de/amm/techinf/layer3/index.html

 

[Toki987]

a logical and learned answer.. Thank you. I appreciate the link as well.
I`d recommend this reading for others as well.

 

[Chibi Nappa]

I don't thing EQ and dynamic compression are really a big part of it..... I'm not sure, but I always though it involved finding out what part of the sound was masked by other parts of the sound, and then removing it. For example, if you have information recorded that describes the 200-700 hz range on guitar 1, but guitar 2 is mostly masking that from your ears, you simply throw that information away. Your ears also need a fraction of a second to recover from loud hits, so information immediatly following a snare hit might be deleated. Stuff like that.