Normalizing vs Compression or Limiting

[barismanco123]

I personally have found that normalizing degrades audio too much and I also learned on my uni dgree that normalising is not such a great thing for audio too.

I thought that compression/limiting is a better way too "close the gap" or relation between RMS and peaks?

Who uses normalizing and for what?

Who thinks it is good/bad or subjective?

Many thanks in advance, Baris

[albee1952]

My 2 cents. I never use normalizing. When mixing, I often use the AudioSuite GAIN plugin to raise/lower the level of bits and pieces to even out a performance, like a vocal phrase sung too soft or a keyboard line or chord played too loud or soft as compared to the rest of the track. I will use compression on most vocals, bass guitar, acoustic guitar and on a stereo buss for drums. When it comes to mastering, I will analyze the entire track to find where the peaks are and IF the track is peaking at lower than -2, I might add gain to get it close to -0(but not all the way). That way, my mastering limiter will have a strong input and I can go very gentle with it(never more than 2db of gain reduction). Bringing in a bounced PT mix that hits -0 tends to sound crunchy and mastering limiters make it worse so if that happens, I will bounce again after pulling all audio tracks down by 2-3db. In any case, I find analyzing and adding gain makes me feel more in control of the final levels(even if that is in my mind). Tried the normalize function in both Wavelab and Pro Tools and just don't like it.

[DrFord]

+1 to Albee.
I have seen on many counts people reporting that normalizing can add errors. Plus, normalizing in theory doesn't help. Making an entire track louder pre-fader or pre-plug ins is silly, when headroom is important. If you need +1db, use the fader. If you need +1db before the fader adjust your output gain. If you need better source gain to print adjust your mic pre, and if you need better singla to your mic pre adjust your mic placement.

Now as Albee said, using the gain plug in to manually even out a recording because of some flubbed note is worthwile. I do the same thing often with vocals particularly because they tend to be more dependant on dynamics = feel. Where as musician's can play louder with the same feel if you need to punch.

[O.G. Killa]

Originally Posted by barismanco123

Who uses normalizing and for what?

Who thinks it is good/bad or subjective?

If you look at normalizing as the process of making your loudest peak, your loudest possible peak, then it isn't really useful. But if you look at normalizing as setting a level, and then all audio gets turned down and/or up to match that level then it can be really useful in certain situations, especially if you can not only normalize by peak value, but by RMS as well.

A good example of RMS normalization is Dolby decoding. Every piece of audio encoded with dolby has a "dialnorm" metadata tag. This tag tells the decoder what the average RMS level of the material is. The decoder then adjusts the output volume to match your current output levels, basically eliminating the need for "volume wars". Ever notice how TV commercials are insanely loud compared to a movie you might be watching on cable? If you run your cable through your audio receiver that has a dolby decoder instead of plugging straight into the TV (and the audio from the station is dolby encoded) , then the commercials will playback through your system at the same level as the movie you were watching.

Another very handy use for normalization is when putting together a library of sounds or sound effects. You recorded them all, but there might be slight variations in volume from day to day... so instead of making them all as loud as possible, you set the normalizer to a median level, that way the few that were a hair too loud get brought down and the few that were a hair too soft get brought up. This way you still maintain the dynamic range and low noise floor that you originally recorded, but have slightly adjusted the files so they are all the same average level.

[barismanco123]

Maybe I have not understood here but you said "If you look at normalizing as the process of making your loudest peak, your loudest possible peak".

The loudest peak will always be the loudest peak and for that reason normailizing on that basis isnt really necersary, no?Or maybe I am not understanding?

I have used the RMS aspect of normailzing in PT and found it to kill the sound even more than in peak mode.

But the RMS version of normalizing in a Dolby system must have a better algorythm and thats why it sounds better as I hear it in the other room now and it seems much better compared to a PT system??? (and I've played around with it in PT - then just though screw it, compress instead).

Normalizing just seems lazy????????????????Just compress instead???

Who uses normalizing without fear of utter distortion or horrid artefacts in PT and forgetting about other platforms - keep it PT centric???

Please help me on my journey of discovery.Baris

[DrFord]

Here's what I think you aren't getting, when you apply normalization, you are simply adding or removing gain from source material until it's loudest peak (in peak mode) is whatever you set.

Ex. loudest peak is -4.1db. You desire to have the loudest peak equal -2.0. (maybe you recorded a stereo signal and your mic pre's weren't set equally, so you want the 2 sides to match up)

Normalizing in peak mode would simply ad +2.1db of gain to your overall region.

In this example no tonal shaping, no coloring, no compression characteristics, no transient modification, no compression whatsoever would happen.

So to answer your question, normalizing has nothing to do with compression or limiting. It is simply a function of gain. However, OG has described the RMS mode, which I will defer back to him on because he knows his sh*t.

We can go round and round on this, but if you want the characteristics of compression, you need to use a compressor. Limiting? Use a compressor on 20:1.

Normalizing isn't lazy, it's a tool to be used for the right function. That function typically has nothing to do with compression, it has to do with adding and removing gain with an automated process. Personally I don't use it. I would rather use my ears and eyes. Plus it can cause errors.

However, it has its uses - again, refer to OG's post.

D

[ChrisMitchell]

Originally Posted by O.G. Killa

Ever notice how TV commercials are insanely loud compared to a movie you might be watching on cable? If you run your cable through your audio receiver that has a dolby decoder instead of plugging straight into the TV (and the audio from the station is dolby encoded) , then the commercials will playback through your system at the same level as the movie you were watching.

Guess what I'm doing tonight when I get home from work??

[DrFord]

Originally Posted by ChrisMitchell

Guess what I'm doing tonight when I get home from work??

The "Love Ewe?" (Inflatable sheep)


couldn't resist. Sorry.

[O.G. Killa]

Originally Posted by DrFord

Here's what I think you aren't getting, when you apply normalization, you are simply adding or removing gain from source material until it's loudest peak (in peak mode) is whatever you set.

Ex. loudest peak is -4.1db. You desire to have the loudest peak equal -2.0. (maybe you recorded a stereo signal and your mic pre's weren't set equally, so you want the 2 sides to match up)

Normalizing in peak mode would simply ad +2.1db of gain to your overall region.

In this example no tonal shaping, no coloring, no compression characteristics, no transient modification, no compression whatsoever would happen.

So to answer your question, normalizing has nothing to do with compression or limiting. It is simply a function of gain. However, OG has described the RMS mode, which I will defer back to him on because he knows his sh*t.

We can go round and round on this, but if you want the characteristics of compression, you need to use a compressor. Limiting? Use a compressor on 20:1.

Normalizing isn't lazy, it's a tool to be used for the right function. That function typically has nothing to do with compression, it has to do with adding and removing gain with an automated process. Personally I don't use it. I would rather use my ears and eyes. Plus it can cause errors.

However, it has its uses - again, refer to OG's post.

D

The Dr is right on the money. Normalizing is not an "algorithm". It is not compression. It is not limiting. It is simply GAIN adjustment across the whole file uniformly.

Normalizing is the process of looking at your audio, determining it's maximum peak value and it's RMS value, and then adjusting it's OVERALL level based on that.

As Dr Ford mentioned... if you recorded a snare drum for example, the first hit reached a max level of -6.2dB, the next one -8.9dB and the third -4.5dB. One use of normalizing is to analyze the recording to find that your loudest peak is -4.5dB. Then to normalize the file to 0dB, the normalize plugin would bring the WHOLE FILE up by 4.5 dB. This would make the first peak -2.3dB, the second 4.4dB and the third 0dB. Or you could say you wanted to normalize the file to -10dB. The normalize plugin would then bring the whole file down by 5.5dB. The first peak would then be -11.7dB, the second -14.4 dB and the third -10dB.

normalizing using RMS is the same thing, except you are using average level instead of peak level. Mastering engineers sometimes do this after they have mastered an album. Every track is peak limited to -0.2dB (usually), but one song might still feel louder than the others... when they analyze it's RMS level they might find it's RMS level is 2 dB louder than every other song on the CD. So they Normalize that one track to the same RMS level as the rest of the CD (and in doing so, make that one track quieter, and the peaks come down as well).

Compression is a different animal. Compression actually makes the peaks softer but leaves the quieter sections uneffected. This allows you to then turn the overall volume up because the peaks are lower, but the "distance" or dynamic range between the peaks and quieter material is now smaller. Normailzing DOES NOT change that relationship/distance/dynamic range, and that's how it differs from compression.

[O.G. Killa]

Originally Posted by DrFord

The "Love Ewe?" (Inflatable sheep)


couldn't resist. Sorry.

And how do you know it's called the "Love Ewe" instead of the Love Sheep??? Speaking from experience? LOL