normalise simply finds the loudest bit of your sample, makes that as loud as possible, and adjust the rest of the samples the same amount.
so if you record something very quiet, you can normalise it to get it to be louder.
there are a lot of things that make it imperfect though for simple 'loudness' boosting - for example, if you raise the entire level you also raise the noise level. also, a very quiet sound with a very short, loud bit (like a record click) will make the normalising not work as well since this loud part is going to hit the ceiling before the quiet sounds do.

By damien907
Tue Sep 11, 2012 2:04 am
ime normalizing works well on things that are about the same level, but if your track has any dynamic range in it, youll **** yourself by normalizing, i just did a couple days ago actually.
the sample i was using had a quiet intro, then the rest of the track dropped and it got louder.
i overlooked this, and when i clicked normalize, it screwed up the quieter intro and brought it to the same level as the rest of the track, but i didnt notice till after i chopped everything and saved it a few times so i coudent undo it.
basically normalize brings the peaks to the loudest they can be without clipping (or the loudest a user defined threshold is set to), thats the easiest way i can explain it anyway.
the sample i was using had a quiet intro, then the rest of the track dropped and it got louder.
i overlooked this, and when i clicked normalize, it screwed up the quieter intro and brought it to the same level as the rest of the track, but i didnt notice till after i chopped everything and saved it a few times so i coudent undo it.
basically normalize brings the peaks to the loudest they can be without clipping (or the loudest a user defined threshold is set to), thats the easiest way i can explain it anyway.
By m56p87
Wed Sep 12, 2012 4:58 pm
Remember: not all normalizing algorithms are created equal.

By tapedeck
Wed Sep 12, 2012 5:10 pm
m56p87 wrote:Remember: not all normalizing algorithms are created equal.
actually, for a true normalising algorithm, yes they are.
it's very basic: find the maximum signal level, divide 1 by this value, then multiply every sample by that value.
Code: Select all
(1/max) * sampleif a normalising algorithm is doing anything other than that, then it is not really normalising.
By m56p87
Wed Sep 12, 2012 6:08 pm
Really? I feel like some are more "destructive" than others. I don't know a lot about code but would it be crazy for software companies to try to improve upon this in minor ways which would could lead to minor sonic changes? Just guessing though, ive heard other people say they prefer certain programs over others for normalizing as well but that dosnt mean shit! 

By tapedeck
Wed Sep 12, 2012 6:30 pm
all of what you say is right - a software company COULD try to improve on a normalisation algorithm - but then it wouldn't be normalisation anymore.
the whole concept of normalisation is because it is a mathematical standard. multiplying by 1/max is how you normalise something.
i think there are perceptive 'loudness' normalisation processes but this is not what i would consider normalisation - more just borrowing the term to describe something else.
the whole concept of normalisation is because it is a mathematical standard. multiplying by 1/max is how you normalise something.
i think there are perceptive 'loudness' normalisation processes but this is not what i would consider normalisation - more just borrowing the term to describe something else.



