does anybody actually use the Export Normalize option?

Discussion of music production, audio, equipment and any related topics, either with or without Ableton Live
mdk
Posts: 914
Joined: Sun Jul 31, 2005 3:51 pm
Location: Skopje, Macedonia
Contact:

Post by mdk » Mon Jan 26, 2009 10:30 am

trevorc wrote:is normalising really doing all that? i thought it was simply bringing the overall volume up until the highest peak touches 0db?
exactly. its not voodoo, its not doing any kind of limiting or compression or fancy trickery.
Pr0k Records - Bandcamp Facebook Twitter

trevorc
Posts: 212
Joined: Thu Oct 18, 2007 11:09 am
Location: valencia, spain
Contact:

Post by trevorc » Mon Jan 26, 2009 10:35 am

mdk wrote:
trevorc wrote:is normalising really doing all that? i thought it was simply bringing the overall volume up until the highest peak touches 0db?
exactly. its not voodoo, its not doing any kind of limiting or compression or fancy trickery.
phew..

bagginz
Posts: 206
Joined: Fri Jan 19, 2007 5:51 pm
Location: Switzerland
Contact:

Post by bagginz » Mon Jan 26, 2009 10:38 am

Yeah, I use export normalize quite a lot. It's a really useful function.

I'm really glad the Abes thought to include it in the export options.

Cheers,
bagginz

anamexis
Posts: 582
Joined: Sat Dec 18, 2004 9:43 pm
Location: Oakland
Contact:

Post by anamexis » Mon Jan 26, 2009 10:40 am

mdk wrote:
blaugruen7 wrote:
Fieldy wrote:but the relation of the gain of the tracks got lost?
yes.
all tracks will be right below clipping.
no, no no no no no.

the *relation* of the gain between tracks will not be lost. i.e. their relative gain will not be changed.
In Fieldy's case, though, he exported every track seperately, in which case normalizing would totally destroy their relative gain.

Machinesworking
Posts: 11081
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Post by Machinesworking » Mon Jan 26, 2009 10:45 am

This does remind me that for throwing crap at the bandmates, it would probably be best to just use a brickwall limiter to get the signal to something close to other songs on their systems.

When you bounce in real time, it decides to normalize in real time too, which is a waste of time IMO. It's useful if you don't want any artifacts in your song. Normalizing really doesn't add that much noise, but raising the volume of any recording will make it seem so.

mdk
Posts: 914
Joined: Sun Jul 31, 2005 3:51 pm
Location: Skopje, Macedonia
Contact:

Post by mdk » Mon Jan 26, 2009 10:49 am

anamexis wrote:In Fieldy's case, though, he exported every track seperately, in which case normalizing would totally destroy their relative gain.
yeah, sorry, i didnt read his first post properly.

although, does it actually normalise each track when you say 'render all tracks' with normalise enabled?

or does it just figure out how to normalise the whole thing and apply that to each track?
Pr0k Records - Bandcamp Facebook Twitter

bagginz
Posts: 206
Joined: Fri Jan 19, 2007 5:51 pm
Location: Switzerland
Contact:

Post by bagginz » Mon Jan 26, 2009 10:53 am

Homebelly wrote:
Any way, what ever it does, it sounds horrible to my ears. A soft limiter , or better yet a, gain plug is a much better idea in my opinion as it will keep a more natural and musicale dynamic..
Homebelly,
with all due respect, you don't understand what nomalizing is.

Normalisation doesn't alter the original sounds in any way other than making it louder by using up any remaining bit depth.

Homebelly wrote:
A soft limiter , or better yet a, gain plug is a much better idea in my opinion as it will keep a more natural and musicale dynamic..
The gain plug would do the same job as normalizing, but that's assuming you boost by exactly the remaining headroom, which is very difficult to know.

Hence we have the normalizing function, which simply lets the computer figure out the remaining headroom precisely and apply the optimum amount of gain necesary.

Using a limiter would change the dynamic range of the original file, and hence by definition, it would be be less natural sounding than the original file.

Cheers,
bagginz
Last edited by bagginz on Mon Jan 26, 2009 10:57 am, edited 1 time in total.

Nokatus
Posts: 1064
Joined: Fri Jul 01, 2005 7:06 am

Post by Nokatus » Mon Jan 26, 2009 10:55 am

Homebelly wrote:I'll never normalize.
I'm not sure how much truth there is in this, but i'm of the opinion that its about one of the worst things you can do to digital audio. Its my understanding that the process looks for the highest peak, compares that to the lowest peak, averages every thing out and then adds the required number of samples to the peaks that fall below, or subtracts them from the peaks that go above.
If you're not sure how much truth there is in that, stop worrying: normalizing doesn't really do that :)

Normalizing checks for the highest peak and adjusts the gain by a constant amount (like a regular volume adjustment) so that the highest peak hits a desired level, in many cases 0 db. That's all.

bagginz
Posts: 206
Joined: Fri Jan 19, 2007 5:51 pm
Location: Switzerland
Contact:

Post by bagginz » Mon Jan 26, 2009 11:01 am

Machinesworking wrote: Normalizing really doesn't add that much noise, but raising the volume of any recording will make it seem so.
Without wishing to be too pedantic, normalizing doesn't add any noise.

The noise flòor to signal peak ratio (signal to noise ratio) remains exactly the same.

cheers,
bagginz

evon
Posts: 1058
Joined: Fri Jun 03, 2005 7:44 pm
Location: Jamrock

Post by evon » Mon Jan 26, 2009 2:29 pm

After reading Bob Katz's book, I try to stay away from normalizing. I dont have the book at hand, but my understanding is that it raises the volume based on some mathematical formula..so there goes the art from all your mixing!

There are all sorts of plugs that one can use to see how many bits are being used and optimizing on the maximum availabe, which is what is desireable at the final mastering stage. Live's meters can also be used as a good enough guide for this also. Otherwise, like some of the other posters have said, I would prefer to work with limiters and compressors and set my master fader to 0 dBFS.

evon
Posts: 1058
Joined: Fri Jun 03, 2005 7:44 pm
Location: Jamrock

Post by evon » Mon Jan 26, 2009 2:35 pm

bagginz wrote:
Machinesworking wrote: Normalizing really doesn't add that much noise, but raising the volume of any recording will make it seem so.
Without wishing to be too pedantic, normalizing doesn't add any noise.

The noise flòor to signal peak ratio (signal to noise ratio) remains exactly the same.

cheers,
bagginz
I should add that there are programs that normalize to RMS instead of Peak. This can be very useful and the outcome is different from peak normalization. The sound wave looks different and the sound is much fuller.

Khazul
Posts: 3185
Joined: Wed Feb 23, 2005 5:19 pm
Location: Reading, UK

Post by Khazul » Mon Jan 26, 2009 3:02 pm

Normalizing is a very simple process - basically you scan the entire track looking for the highest peak individual absolute sample value.

For eg... with sample values of 1,337,2443,234,3,-257,-763,-3245,-891,-246,3 ...

Then the highest absolute (ie furthest from zero) value is -3245. With a 16 bit respresentation as per CD, then the possible range of values is -32767 to 32767 (or it might be 0-65535, but lets assume the former, and lets assume the max -ve is -32767 instead of -32768).

So normalizing will work out how much gain you need to raise -3245 to a value of -32767 ie -32768 / -3245 which is about 10.098. So effectively to raise thje gain of all samples by multiplying them by 10.098... in this case.


The level of music audio has two important aspects:
1 - how loud it sounds on average
2 - what the actual peak sample value is.

Dynamic range of music can be said to be the difference between its average level and the peaks - anyone who using K-System metering with both VU and peak levels already probably get this.

You can have two tracks with exactly the same music at exactly the same apparent average level, but where in one case, some peaks, for example where a kick and snare and a hi-hat hit exactly together cuase huge peaks, whereas in the other a brick wall limiter has hugely reduced those peaks.

Or more likely a case where you have used two different kick drums - one happens to have a huge initial transient and the other doesnt, but they both basically sound the same.

Normalizing something with huge peaks is going to produce a way smaller level increase than normalizing something that doesnt have huge peaks - this is why normalizing for the sake of it is pretty pointless.

When to normalize - after running thorugh a limiter to get the desired dynamic range - the normalizing will bring the level up to the desired level. In this case normalize to a tiny bit below 0dB, for eg -0.3dB.

When loading a sample into a sampler and you dont care about that sample's current relative level with other samples, however you do want that sample to make the best use of the dynamic range of the sampler - for eg 16 bit sample that had to be down converted to say 12 bit to load it into an Elektron machine drum, so you are trying to make the most of the available dynamic range in the sampler.

Why to avoid normalizing anything whose entire signal chain is not known to 0dB (ie you didnt record it, it came from a sa,mple library etc)? - if (for eg) a kick sample has allready been brick wall limited or worst, simply clipped by a cheap brick wall limiter approach, then you may end up with consective samples at or near peak level. When that clipped waveform is reconstructed (in the analog domain) then it can result in values above 0BFS - ie it could actually clip. Therefore when normalizing potentially clipped samples, dont normalise any higher than say -3dB, this will give you some room back for the waveform reconstruction to occure without clipping. Note - this isnt just at the point of use, but anywhere is the subsequent processing chain - ie dont let you kick drum sample hit full scale when you limit your track - it might actually distrort horribly on some equipement and not others.

In modern production the two point of normalizing are usally down sampling - the machine drum example above to maximise available dynamic range and the implie normalization that tends to occure as a consequence of settig a brick wall limiter's outpur level on you master bus to -0.3dbFS while your master bus is at zero. Effectively the limiter outputs a signal normalized to whatever its output level is set to.

Hope that makes seome sense.

IMHO in a daw its not a function that you will generally need to use, or at least not on its own.
Last edited by Khazul on Mon Jan 26, 2009 3:03 pm, edited 1 time in total.
Nothing to see here - move along!

mdk
Posts: 914
Joined: Sun Jul 31, 2005 3:51 pm
Location: Skopje, Macedonia
Contact:

Post by mdk » Mon Jan 26, 2009 3:03 pm

evon wrote:After reading Bob Katz's book, I try to stay away from normalizing. I dont have the book at hand, but my understanding is that it raises the volume based on some mathematical formula..so there goes the art from all your mixing!
you're right that it uses 'some mathematical formula'

A = B * C

where
A is the final audio
B is the audio on the master
C is the gain required to make B have a peak level at zero

who knows what maths like that will do to our precious art?
Pr0k Records - Bandcamp Facebook Twitter

leedsquietman
Posts: 6659
Joined: Sun Nov 19, 2006 1:56 am
Location: greater toronto area

Post by leedsquietman » Mon Jan 26, 2009 7:04 pm

agree with khazul again

Normalizing sends the peak to what it sees as 0dbFS but it is not very sensitive and does not take into account possible intersample clipping and on material with very fast attack transients, might overshoot 0dB, causing clipping.

If you're using 16 bit, I could see why people do it, as there is more poitential noise if you don't get levels a bit hotter, but in 24 bit with all the headroom you have, there's no need. Even in 16 bit, it would help if the normalize feature was set to something a bit lower than 0dbFS, to prevent intersample clipping. A good limiter like Voxengo elephant has 8x oversampling which really helps to eliminate intersample peaks/clipping. Live's normalize feature does not. Besides all of this, it's a good habit to get to know how to do gain staging and using compression and limiting without using a one click normalize button.
http://soundcloud.com/umbriel-rising http://www.myspace.com/leedsquietmandemos Live 7.0.18 SUITE, Cubase 5.5.2], Soundforge 9, Dell XPS M1530, 2.2 Ghz C2D, 4GB, Vista Ult SP2, legit plugins a plenty, Alesis IO14.

laird
Posts: 1996
Joined: Tue Feb 27, 2007 10:05 pm
Location: PDX
Contact:

Post by laird » Mon Jan 26, 2009 7:21 pm

mdk wrote:
trevorc wrote:is normalising really doing all that? i thought it was simply bringing the overall volume up until the highest peak touches 0db?
exactly. its not voodoo, its not doing any kind of limiting or compression or fancy trickery.
Interesting side note: Live's normalize brings the overall volume up.... or DOWN... to make the peak hit 0.0 dB.

As for how much noise it adds, etc... use it if you need to (like when you want to export a ton of samples without having to mess with the volume each time)... and don't worry so much about the noise. I bet bad mixing skills and poor EQ choices or brickwall limiters add more "noise" to your mixes anyway ;)

Post Reply