Normalizing suggests that you can make a group of sources all the same volume. In the whole technical sense,
a Normalizer will do that. It will increase the loudest signal in a sourse to 0db on a digital recording, the highest you can go.
Many people believe that by Normalizing, they are getting the most volume possible out of the track. But what they are really getting is the maximum PEAK volume out of the track, and this has nothing to do with average volume, and has nothing to do at all with making a track with a lot of low and high frequency sound as loud as a track with a lot of midrange information at the same average volume because the ear perceives different frequencies to be different volumes, even though the have the same sound pressure level in real life. The ear just does this and there is no way to changed your ears (except for the worse....
)
If all you want to do is make sure that your track's highest volume actually reach's 0db, then normalization will do that. But, you may never really perceive this increase in volume because the audio track could have transient signals (very fast jumps in the volume of the audio, usually very hard to hear high frequencies) that can be up to 6db or more louder than any other part of the audio.
Let's say that your audio has one transient that peaks at -1db on the recording. A normalizer would detect this and would turn up all the audio 1 db. Hardly a noticable difference at all!!! What have you gain from the processing? Nothing really. You have done little to improve the average level of the music which can be widely different.
Now, let's say you have another track that has hardly any difference between the PEAK level and the average level. The Peak level is at -6db, and it is
only 1 db higher than most of the rest of the track. By applying a normalizer, you would get an over all +6 db gain out of the average level! +6db is a lot of increase. Technically, that is twice as much sound pressure level (db).
So, you have increased a track with very high transients but very low average volume, and a track with little to no transients and very low average volume. Let's even say that the average volume of both tracks is about the same. The track with less transients will enjoy a hugh increase in volume while the track with many hot transients will get little benefit. So in effect, you really haven't "normalized" anything. You just increased the Peak level of the recording to the highest level possible.
People talk about using a Normalizer plugin for mastering their recordings. This is not a very good idea at all. Mainly, it does not really work except in the crudest ways. It is seldom what the person is actually after, which is a better average level for the tracks.
Part of the job of mastering is to make each song to be "perceived" as being the same volume. To accomplish this, mastering engineers have to make many decisions about how they will make a song that contails a lot of transients, but low average level sound as loud as a song that has little difference between the highest transient and the average volume. Generally, the mastering engineer is going to use some kind of compression or limiting on the song with hot transients to that the over all level can be brought up without creating digital distortion due to the transients being brought up too much. So the limiter would be set to only engage at a level just above the average volume level of the song, thus, the transients would be the only parts that would get limited. Then, the make up gain on the limiter can be turned up to make the overall volume of the song louder because the gain reduction would tame the hotter signals only, thus making more room for a overall volume increase.
Their are problems with doing this though, depending upon how much higher the loudest peak is over the average problem. Too much compression and/or limiting can create distortion. It can also tend to make the song sound dull and lifeless. It can in effect make the audio sound to muddy.
Transients are a normal part of music. It is when the transients are too far above the average volume that they become a problem. This holds true regardless of whether you are recording analog or digial. With digial, overly hot transients make your average record level come down quite a bit, thus, you use less bit on the A/D/A converters, thus creating a boxy, kind of edgy, dull digital sound that everyone seems to complain about. With analog, transients would cause lower record levels too, but the adverse effects are tape his and equipment noise. Dolby helps some, but not totally. All in all though, overly hot transients just make the recording too low in volume. Not really a problem untill you start comparing your mixes to mixes where the transients where tamed properly.
Most people are looking to get an overall volume increase in their recordings. Normalization can help, but their is one other little problem with using it after a track has been recorded.
Normalization cannot make up for the lack of detail in the sound that is the result of low bit resolution. Well you record something too low in digital, the sound will lack much of the original detail because the A/D converters did not get enough volume to offer enough data on the recording. So, normalization just basically turns up the already bad sounding audio. Since digital enjoys a pretty damn good sound to noise ratio, this is not really a big deal. It is not like a digial mixer is going to add more equipment noise because you had to turn a low volume track up a bit to get it as loud as you need it in the mix. So really, the gains of applying a normalizer on an audio track are minimal.
The last and final point that needs to be addressed about this whole thing is relative loudness.
Like I said earlier, mastering is where you usually make all the songs of a CD "seem" to be the same volume.
If you where to analyse many recordings, you will find that they do not sport the same average volume. They can differ widely from song to song. Why? you ask. It has to do with the relative perceived volume of frequencies to the human ear.
I have suggested many times that anybody audio engineer become very familiar with the Fletcher/Munson Relative Loudness Curve. These two scientist provide a very detailed and usable graph of how the human ear perceives the sound equal sound pressure levels of different frequencies.
Without going into too much detail here ( I am trying to get you all to read it damnit!!!
) the relative loudness curve suggests that the human ear hears midrange frequencies much better than low and high frequencies. The Fletcher Munson Relative Loudness Curve shows a graph of the relationship of this phenonomon based upon a 1 kHz test standard. Basically, it show how much louder a certain frequency needs to be to be perceived as being as loud as 1kHz at any certain volume.
It also shows that as audio starts getting louder, the difference between the perceived volumes of frequencies starts to even out. But at low volumes, the low and high frequencies need to be much louder than 1kHz does to sound the same volume.
I am assuming that you are all smart enough to figure out how this could apply while mixing music. It REALLY applies to mastering because one of the aims in mastering is to make all the songs sound about the same volume. Some songs have more midrange information then others, so these songs will tend to sound louder at the same average volume. Usually, the song with more midrange info needs to be turned down a db or two to make it seems the same volume as a song with more low and high end information, especially in the case where the average level of the song is pretty much as loud as it can get without distoring because of the transients on the track.
So, study up guys. Some of you may think that this kind of stuff is not relavent to producing clean