Does Normalizing affect sound quality?

I completely understand that there is a lot to good mastering, and Normalizing would not be part of it. Having said that, I want to know if normalizing would ever have any affect on sound quality. I am looking for a way to set bounce levels for a project so i can get my stereo file to start mastering. It seems like normalizing the project bounce could be an easier way to achieve this than searching for a peak. Again, this would be when bouncing down the project to the stereo interleaved to use for mastering.

 

Thanks.

Normalizing can indeed be a part of mastering, and might be the final step. But if you are exporting tracks to be mastered, it would imply there would be no headroom for the mastering engineer to work with (they could just drop the level of the tracks to provide their own headroom, I guess). Still, I don't really see an advantage one way or another. It doesn't really hurt, but it doesn't really help in any way.

 

Normalizing never affects sound quality. All it does is identify the digital bit in the track that has the highest value below 0 dBFS, calculate the difference between that value and 0 dBFS, then add that value to every sample. This is really just turning up the volume correspondingly the same on every bit, except permanently in the print.

 

This means that the relative difference between bits stays the same, and the relative level of the noise floor also stays the same, even if everything has gone up a dB or two in printed level. So that means that nothing in the sound quality has changed, and the only thing that has changed is the output level of the track.

 

Since that level is then manipulated during mixing, it really doesn't matter whether it has been normalized or not. It will either be turned up or down accordingly in the strip anyway. And since it is not really an algorithmic process with rounding, it does not chain more rounding error into the track. For tracks that will be mixed, normalized vs not normalized is functionally the same.

 

One advantage of normalizing a single track inside a project is so that the level is higher, but only so that a noise gate can be used to lower the corresponding noise floor between notes back to where it was before normalizing (as the mixer turns the entire volume of that track down). This means the relative noise floor is lower during rests, and still masked during notes.

 

Since we also want everything to have as high a level as possible in the final exported track, it also makes sense to normalize there during export, although that will not change anything regarding the actual quality, neither pro nor con. We want one sample to touch 0 dBFS. Just one.

 

But I can think of no other reason where normalization could benefit anything, other than sometimes you can raise or lower (or normalize) an individual note in an audio track, which is sort of like destructive automation for a single note.

I wonder what happen to inter-sample peaks, which could potentially create distortion, which normalizing algorythm might be ignoring

Thanks. I got the sound quality answer. Not to open a canoworms, but I'm curious as to the reference of normalizing as last step of mastering. I'm ending with the Adaptive Limiter.

I don't normalize to prepare for mastering, and I don't normalize as the last step of mastering. 

As previous poster(s) mentioned, normalization to 0 dB means you have to deal with inter-sample peaks; it's bad gain staging, and there's really no reason to make another copy/bounce of a file normalized to 0 dB.  Playing back audio normalized to 0 dbFS can add distortion and so the general rule of thumb is to limit  peaks to -0.3 dB at then of the mastering chain, even when using a limiter.  Playback of a file normalized to 0 dbFS using a limiter with inter-sample peak detection turned on (with the output set to 0 dbFS) sounds worse compared to playback of the same audio with peaks at -0.3 dBFS (with inter-sample peak detection turned off).   So normalizing to 0 dbFS is usually first on the list of things never to do - followed by adding too much compression or limiting in your final mix.  Force yourself to make at least one mix with no compression and limiting on the master buss; most mastering engineers will pick the un-normalized, uncompressed and unlimited mix if they have a choice.

 

My experience is mastering engineers want 6 to 12 dB of headroom to work with.  That would mean to at least avoid peaks above -6 dbFS in the final mix.  With 24 or 32 bit audio files, mixing peaks to -6 dbFS or even lower is not going to hurt the signal to noise ratio.  But if you normalize peaks to 0 dbFS, then apart from inter-sample peak/distortion issues, you just end up forcing the mastering engineer to reduce the level back down from 0 dbFS by however many dB's of dynamics he is trying to add via EQ cuts, expansion, etc = poor gain staging.  (And then there is the distortion added by any compression and limiting from the master mix buss).  Forcing someone to master with a file normalized to 0 dbFS - where 6 dB of dynamics will ultimately be added by mastering - is unlikely to sound as good compared to starting with a mix with peaks at -6 dbFS in the first place.

 

While it's not true for all kinds of music, a "crest factor" of 12 - 15 dB  (the average difference between the peak and RMS level for the entire track) is a good rule of thumb for a final master (with peaks not exceeding -0.3 dB).  If your mix can already produce a ~10 dB crest factor, then probably no need to mix peaks to less than -6 dbFS before mastering.