Mp3 faq - normalization

What does "normalization" mean? What's the difference between normalizing "peak" level and "average" level?

If you say:

"Normalize to 98%, but only if peak level is higher than 98% or lower than 90%"

This means than if the highest sound peak (sample amplitude) in the file is less than 98% and more than 90% then nothing is done, but if the highest amplitude is more than 98% or less than 90% then the file will be adjusted such that the highest amplitude will be 98%.

The other options mean the same thing, but based on the *average* level rather than the peak level. Generally the average level will have a closer connection to how "loud" the track sounds than the peak level will, so you will probably be better off normalizing based on the average level... but if you normalize the average level you may cause the overall level to be increased to such a degree than some "clipping" will occur (i.e. the conversion will try to produce levels higer than the maximum possible, so the peaks will be "clipped" off). This is all probably a bit clearer if you use a WAV editing program where you can actually see a graphical display of the file, showing the level as it varies over the length (time) of the file. There is often no perfect answer to the normalization issue. Theoretically, it might seem best to normalize to a low average level - that way you're mostly going to push files down in level rather than up, so you shouldn't cause clipping, and you'l
l end up with a bunch of uniformly fairly soft songs, so you'll have to play back with your volume turned up rather. To be honest, I never actually do this.

For classical music, or other music where very soft passages and large changes in dynamics are important, I'd go through all the tracks to be put together, and check their maximum levels - if none of them has a peak of 100%, then adjust the one with the highest level to have a peak of 100%, and then adjust all the others up by the same ratio (i.e. for each make the new peak op * 100 / omp, where op is the old peak of this file and omp is the old peak of the file with the maximum peak which you adjusted first). If one or more tracks already has a peak of 100%, I'd leave them alone.

For rock music, I tend to take a different approach. Basically I increase the average levels of softer tracks, not to make them equal to the average level of the loudest tracks, but to produce a subjective listening level which seems "in the same range". You can only really tell by listening. And yes, I probably cause some clipping when I do this, but quite honestly I don't think a little bit of clipping here and there is a big problem with rock tracks, in general.

top of the page