And the problem with normalizing is?

[tsw]

I know normalizing is about as popular as Barry Manilow. I want to know why. Since I couldn?t afford a Masterlink, I got Bias Peak (not bad for $99!) for my G4 iBook. The normalizing function lets me get better levels without changing the proportions of the dynamics of the song. The quiet parts are still proportionally quiet.

How is that a bad thing? I?m not trying to start an argument here. I?m really asking, what is the complaint against this function.

andy

[tiger vomitt]

it's not bad at all, as an operation done by itself.

if you're going to do other stuff to the file after normalizing, then you'll run into problems.

i like Peak too. been using it for years, good app.

[joel hamilton]

Who said it sucks?

It works. it is one way to get things louder without destroying the relative amplitide information in any track.

Pop engineers normalize vocals all the time after editing. they still compress them outside the box, but normalizing can give you a more "up front" feeling track just by bringing the loud bits right up to "0" without any overs.

Not liking normalizing is like not liking editing, or mixing, or compression, or reverb.

It is a tool that can be really handy under the right circumstances.

[@?,*???&?]

Joel takes a pretty neutral position on this and I'd have to err on the negative side. I haven't heard the imperative aspects of normalizing. I much prefer the multi-stage compression or limiting done while mixing. When comping a vocal, I'll often ride levels of the vocal to make sure it's more levelled when assembling that master track. Normalizing seems like a one-step band-aid applied at the end of the process. Why not just use an L1 on the vocal and be done with it? You get my point...

[joel hamilton]

Jeff,

I totally respect your opinions, and I have learned things from reading your posts, BUT:

L1 will limit something. Normalizing is not limiting. The relative amplitude within any given selection will stay intact when you normalaize, which CANT happen when compressing and or limiting.

in other words, if you want to keep your dynamics intact in the digital domain, but bring up the levels in a proportionate way, then normalize and use some real stuff outside the box.

if you want the loudest sections to get clipped, and the quieter parts to be louder: L1.

When you normalize, quiet is still quiet, and loud is still loud, but both are louder.

Ya knowwuddimsayin?

The closest thing to normalization in the analog world is YOU going through and finding the loudest part of a track, setting the levels accordingly, and going back and reprinting the track to another track with NO dynamics in the re-print chain.

Horrible sentence, but I think you can dig it.

[@?,*???&?]

Joel, what's the end result of 'normalizing' a track that has already been normalized? Can you 'normalize' a track multiple times? What would that achieve?

[Scodiddly]

Normalizing is basically just a way of getting the levels up to the exact maximum level available. So if your track has the loudest peak at something like -6dB, the normalizing software will mathematically increase the entire track by 6dB so that the loudest peak is at 0dB. This is a more precise (and because done in software, easier) version of very carefully watching the meters and setting the record level just exactly right.

The downside is that it's a bit of math applied to a digital signal, so you can get rounding errors and such. If you're still at greater than 16 bits of resolution then it's not a problem.

Doing it more than once will have the same effect as having done it only once.

[spiral]

I've never understood what the opposition to normalization was. It is just turning up the gain basically, without clipping. Even if it was destructive to the original audio (and i'm not saying it isn't. i just don't understand why.) it can't be as destructive as compression.

[Rigsby]

I normalise almost as a matter of course to get things up to a similar level after mixing, so i take the whole mix and normalise that. Are there problems caused by this? You have me a little worried now. Is the sound affected, as it doesn't seem to be?

Oh dear, what have i done now?

[marqueemoon]

For a final 2 track mix, after you have done everything else you are going to do with it, it's fine. The problems really crop up when you're working with mixing a set of normalized tracks, as you can run out of headroom more easily.

[Russian Recording]

Normalizing isn't necessarily good or bad, it's just a tool. What normalizing does is it simply finds the difference bewteen the loudest peak and 0dBfs in a given selection, and raises the gain by that amount so that the selection is as loud as it can be without clipping out.

The problem is that the peak does not say anything about the average loudness of a song. For example, let's say you have two songs that are going to be on an album. One is a heavy rock tune with full band, the other is a slow tune with piano and vox only. The really quiet song is not going to have very drastic peaks whereas the rock song is going to have plenty from the drums (mostly the snare and kick). When you normalize the two, you will see (hear) that the quiet tune is actually way louder than the rock tune! The reason is because there are no fast transients in the slow tune so the average loudness is much closer to the peak and therefore can be brought much closer to 0 dBfs. With the rock song, there are a lot of fast transients from the drums and therefore the average loudness of the song is much further from the peak.

In order to make the rock song seem as loud, or louder than the quiet song, you can do one of two things.

A) You can turn down the quiet song (ie not normalize it) to a level where it feels right in terms of average loudness compared to the rock song (this level can only be achieved by listening, although using VU meters can be helpful [VU meters are designed to represent the average louddness of the source material rather than peak meters]). In a world of 24 bit recording, one does not need to worry about getting down into the "nasty/noisy" bits unless you have to turn it down by 42 dB, which is most unlikely. Also, by not normalizing, you are leaving some "room/bits" up top, which is useful for the mastering engineer so he has the option of turning the song up or down, rather than just down and also gives him some extra bits/room for additional digital processing such as EQ and dynamics (every digital process, including gain changes, creates a larger wordlength, which is why digital processors and DAWs have internal word lengths of up to 64 Bits).

Now the other alternative....
B) Unfortunately these days, most engineers don't even realize that option "A" is even an option. So, the other option is to keep the quiet song maxed out (normalized), and squash the fuck out of the rock song to not only make it as loud, but even louder than the quiet song. So now all of those nice transients and peaks need to be compressed or limited so that the peaks are much closer in amplitude to the other stuff (the less transient instruments like the bass, guitars, vox, etc) . This is one reason to compress/limit snares and kicks while mixing, because it can prevent one single extreme transient from keeping your average loudness too low. If you have to compress and limit, it is much better to do it on individual tracks during mix down rather than to the final 2-track mix. If all you have now is the 2-track mix, you are forced to compress and limit the overall mix, which can dramatically affect the entire mix rather than just the transients.

This how a simple tool like normalizing can lead to the disaster that we now know as hyper-compression. It eventually gets to the point where you physically cannot get any louder. 0 dBfs is 0dBfs, period. As a result, every song is completely obliterated until there are no transients or dynamics at all. Every song is the exact same loudness, and every song sounds like absolute crap becasue there are no transients and no dynamics not only within the entire record, but within every second of every song. It has gotten to the point where there is audible clipping throught an entire records... even on the quiet songs... and Im talking about major label releases. Listen to the Zwan record, or RHCP's Californication, or At the Drive In's last record. It's a fucking disgrace, I cant evenlisten to this stuff anymore cause it physically hurts my ears by the time I get to track 3.

Ok, this turned from an informational post to a rant so Ill stop now. I hope the explanation above is not confusing and is helpful in some way. It's hard to explain stuff in writing.



mtar

[vvv]

To say pretty much what has been said above, but maybe in a way somebody might find easier to understand (maybe not! ):

if you are normalizing your mixed tracks (the completed songs) to each other prior to going to CD, you are basically lowering everything, or a lot of the tracks, to the average of all of them; e.g 50% will be lowered, 50% raised. This is where most people object to "normalise"; it is handy for quick "mix" tapes/CD's but should not be done when presenting an artist's work as it is an inefficient, albeit convenient, shortcut.

Note also that where "normalize" on an individual raises the track based upon its peak, compression and/or limiting will (should, if intended to) take down those peaks, allowing the average level to be increased.

[spiral]

I can understand how normalization can be "dangerous" as a substitute for proper mastering, but seems like just the thing to do on a quiet vocal track etc.

[joel hamilton]

To say that normalization during the mix eats headroom is like sying pushing all the faders up during mixing eats headroom.

Normalization, in practical terms, is a "smart" gain change. Yes, it is a processed gain change, which could be debated as to the benefits, but it is a gain change nonetheless.

If you track something with transients that hit 0, and you normalize to 0 you will see/hear no change other than file name.

if you record a snare track with only ONE hit that reaches zero, same deal.

Another example:

if you have 4 snare hits, and one of them reaches -3 and the others are all at -5. after normalizing to 0 you would have three hits at -2 and one at 0 (if 0 is the threshold YOU have set).

Normalization is a processed fader move. If you see that your mix is hitting -3 at the two track, and you push the master fader up to make the entire mix hit 0, you will still retain the relative volume information contained within the track, but you made ALL that info come up a bit.

I cant seem to make a simple statement that clearly illustrates this function today.

/babble

[Russian Recording]

I can understand how normalization can be "dangerous" as a substitute for proper mastering, but seems like just the thing to do on a quiet vocal track etc.

Yes, it can be useful for some stuff. It is a nice feature on DAWs when tracking/mixing if you want to turn something up quickly, or if a great take was recorded absurdly quietly. However, by bringing everything up to 0dBfs, you are running the risk of truncation and/or clipping when using additional effects/digital processing. If you get good levels to "tape" there really is no reason to normalize, it's not worth an extra 2-3 dB, and it doesn't improve the sound in any way.
mtar