Arash Ahmadi wrote: ↑
Fri Dec 07, 2018 4:16 pm
I'm not talking about Iphone normalization. I do it on Cubase 9.5 and have tried it on Wavelab 9 too. I am not a pro sound engineer, I'm self taught. How do you define normalizing? How do you do it?
OK. Simple (or 'peak') normalization just means to multiply all the amplitude values in a digital waveform representation of an audio so that the max value in the sound file is bumped up to be the max value possible. All amplitudes are multiplied by the same factor (which is essentially an amplification).
1. How can this go wrong?
This should be done on a file which was either uncompressed(*) (like wav), or losslessly compressed (like flac) - the program in either case would uncompress it - and the result should be saved without lossy compression as well.
The program you are using should be resampling the file at a high rate so that the (floating-point) multiplication can be done with greater precision, preserving better the original shape of the waveform, i.e. reducing artifacts from the multiplication itself.
If you are hearing audible differences in sound quality just from simple normalization, check if the procedure is as described above.
2. What if there is no difference in the volume after simple normalization?
This can happen, because there could be a single (or just a couple) of amplitude values in the audio file that are already at or near the maximum allowed value (we can't really perceive single max values as loud so they may not be noticeable from listening). So the multiplication factor is 1.0 or very close to it and nothing much happens, meaning the input and output are identical.
So normalization then can include a bit of other manipulation like dynamic compression in conjunction with trying to perceive the loudness as a human would, so it runs through the file with a 50ms amplitude-averaging window or something, and then uses these averaged values in order to calculate the multiplication factor (the amplification). Because we can't have individual amplitudes beyond the max value, it must do some compression to reduce those individual values while it is doing that.
(*) 'un/compressed' in this context means compressed for file size, NOT referring to dynamic compression which compresses the range of sound amplitudes in the file.