What's the real story with "Normalization?"

[TBman]

Some people recommend it, some hate it and say do something else to boost the sound file. What is the general consensus on its use?

[pieterh]

As I understand it normalisation brings the highest peak level of a digital recording (eg an audio track in Logic, Reaper or ProTools etc) up to 0dB (or a lower specified level).

It isn’t really needed on digital multi tracks as being digital you don’t really gain anything though there is no tape noise or system noise that gets amplified too.

Recording digitally there is a good reason to have a bit of leeway in the recording levels (tape saturation can be a good thing, digital not so much!) but if the levels are too quiet then you can be struggling to bring up a quiet channel in the final mix.

I used to find Reaper (or my interface) would record at relatively low levels and would normalise some tracks just to make them easier to work on. The downside is that Reaper normalises to 0dB with the result that as soon as I added anything (boosted an eq band etc) then the channel would go into the red.

In Logic you specify a level in percentage terms which allows you to bring working levels up without using up all the headroom before you’ve started.

Part of the mastering process as I understand it is to bring the peak levels of the final mix to 0dB. However unless you are doing the final mastering I think the recommendation is to leave some headroom for the mastering engineers to work with.

[jim1960]

Normalization is simply changing the volume to reach some specific target level. Unfortunately, there's never been a real industry standard in the digital age and that has resulted in what is referred to as the "Loudness War."

[Brent Hahn]

Originally Posted by TBman

Some people recommend it, some hate it and say do something else to boost the sound file. What is the general consensus on its use?

I don't know what the general consensus is, but it does no harm. Anyone who thinks they hear a qualitative difference other than volume is mistaken.

[Joseph Hanna]

Originally Posted by jim1960

Normalization is simply changing the volume to reach some specific target level. Unfortunately, there's never been a real industry standard in the digital age and that has resulted in what is referred to as the "Loudness War."

Well.....not exactly Jim Most simple normalization functions in the DAW’s most of us use defaults to “peak normalization”. That is to say it is a look ahead algorithm that analyzes the current peaks and raises those peaks to a pre-determined level, usually zero dBFS. The important note here is all of the original dynamics remain relative. Where that indeed can raise over-all levels it (normalization) doesn’t have a major stake in what is commonly referred to as the Loudness War.

The loudness wars are largely a result of what’s commonly described as “perceived” volume. Perceived loudness is generally a result of contemporary software known (at least originally) as bit-depth maximizers which is much more akin to brick-wall compression. Brick wall compression, be it a digital algorithm or old school analog device, does mainly the polar opposite of normalization in that it aggressively changes the relation of the dynamics. Once those dynamic differences have been altered the algorithm then raises the peak to a pre-determined level. That in and of itself is primarily the king pin of the loudness war.

[Gtrfinger]

I wasn't aware that normalisation actually increased the volume at all, just made the wave forms bigger so they're easier to work with.

[runamuck]

The simplest explanation I can think of is that normalization makes a recording as loud as it can possibly be without distortion.

[muscmp]

no real consensus that i've heard. use it if need be, don't if it don't. it is an individual thing.

play music!

[rick-slo]

I don't use normalization. As already posted, it adjusts the loudest point in the recording to some small degree (you can set the amount) below the digital max possible without overs.


Problem is that before as well as after normalization different recordings have different dynamics and "average" volumes. So a group of normalized recordings
can vary quite a bit in volume. For a group of recordings to work together as a listening experience I suggest setting volume gain (or reduction) of each recording by ear.

[Joseph Hanna]

Originally Posted by muscmp

no real consensus that i've heard. use it if need be, don't if it don't. it is an individual thing.

play music!

Going back a few years, which is now-a-days what I do best, AND assuming we’re talking about normalization used on a finished stereo file, the only argument would have been the algorithm itself might introduce some sonic anomalies. We argued for what seemed like centuries over which dithering algorithm sounded best so it makes sense we argued about normalizing algorithms as well.

On the other hand and for those relatively new to recording, normalizing “individual” tracks (especially to zero dBFS) is utterly bound to rear many ugly heads as a multi-track project moves forward.

I see little or no down-side to using normalization once understood.

[Mr. Jelly]

Originally Posted by rick-slo

I don't use normalization. As already posted, it adjusts the loudest point in the recording to some small degree (you can set the amount) below the digital max possible without overs.

Problem is that before as well as after normalization different recordings have different dynamics and "average" volumes. So a group of normalized recordings
can vary quite a bit in volume. For a group of recordings to work together as a listening experience I suggest setting volume gain (or reduction) of each recording by ear.

This is exactly what I found. To my mind it doesn't work for anything that I can think of. But to each his own.

[Joseph Hanna]

Originally Posted by Mr. Jelly

This is exactly what I found. To my mind it doesn't work for anything that I can think of. But to each his own.

Hypothetically. You have a finished stereo interleaved file. You’re looking at a meter and noticing the peaks are roughly -12 dBFS. Perhaps you’re sending this file to a friend or posting to SoundCloud or YouTube or Vimeo. Or conversely you’re sending a file out to be reviewed by a client and you know the client is gonna listen on their MacBook.

Maybe, as is now becoming the norm, the project is gonna be exclusively social media based but you still want to retain the massive dynamics you’ve worked so hard to create, yet still be viable on ear-buds.

12 dB of overall volume gain without distortion while simultaneously maintaining dynamic relation is a pretty darn worthily positive, especially in this day and age of the other guy is loud.

[Bob Womack]

Here's what normalization does: You set a distance between absolute saturation ("zero," ie. 100%) and the final peak level you want for your loudest peak. I work in Nuendo. In Nuendo we've discovered that it is possible to actually cause distortion at the peaks of a file you've asked to be normalized to zero. To prevent that, in my house we choose -.1 as our normalization level. The program simply grabs the highest peak in the whole clip and draws the volume of the whole file up, without disturbing the dynamic structure of the recording, to the point where that peak matches the level I've selected. Folks, that's all there is to it. It is a tool.

Here is an example of what it is useful for: Nuendo only offers twenty-four db of gain above zero on clip gain. On the positive side of gain, the automation on a track goes from zero to plus six db. If someone sends me a clip with an average volume of -35 and I need a portion of the clip to be at zero, no amount of cranking up the clip and automation gains will bring it up there. I need to normalize the clip and then start from there.

Bob

[Brent Hahn]

Originally Posted by Bob Womack

In Nuendo we've discovered that it is possible to actually cause distortion at the peaks of a file you've asked to be normalized to zero.

Is that true if you export a full-level normalized WAV and play it back with a different DAW and converter?

Or, put another way, is the distortion actually in the Nuendo digital file, or do you hear distortion because the analog DAC outputs can't cleanly reproduce that much level?

Or, put a third way, if you do a full-level normalization in Nuendo and the file sounds distorted, if you then do another normalization of that normalized file, but this time make it peak at, say, -10 dB, does it still sound distorted?

[rick-slo]

I would think normalizing to 0, even with high upsampling, is bound to let data through that will create
distortion above the Nyquist frequency which will be aliased back into audible frequencies.