Are You Giving All the Right Signals?

[aj113]

Another point which has got me thinking recently:

My current method of mastering is to import the rendered stereo track into Soundforge and then normalise to 0db. From there I have a known quantified platform (i.e. a wav normalised to 0db) from which to operate my limiter.

Now, I know many of you cut out this step and master realtime from the DAW with the limiter plugged in, but there two downsides to this for me:

1. You can't see the resultant waveform either before or after the limiter - both are important to me.
2. You can't be sure that the signal before the limiter is normalised to 0db. It could be anything, you can't possibly know. I know some of you will be saying 'so what, the limiter will take it to the required loudness and we can check the rms rating realtime so it really doesn't matter what the db of the original signal is.

I am not so sure. I am of the mind that a low signal boosted radically by a limiter to achieve the required loudness level is not the same aurally and sonically as a 0db signal boosted less by the same limiter to achieve the same loudness level. For certain no-one who masters in this way can talk with any confidence about the degree of db boost that has been applied, because they don't know what their starting point is.

If that is the case, then surely we want our original source file to be optimised to 0db first before applying any limiting?

Or maybe I'm missing something? I'm sure there are some learned people here who know a lot more about the technicalities of this than I do.

[acidfrost]

Like you mentioned, I'm one of those who "master" in realtime as it's part of my mix and sound.

1. You can't see the resultant waveform either before or after the limiter - both are important to me.

I'm curious about why it is important to see it?

You can't be sure that the signal before the limiter is normalised to 0db.

The point of the limiter is to bring it to 0db (or in my case I put -0.1db just to be sure in case the limiter misses something)
That's how you know your final mix hit 0db is with the limiter.

I am not so sure. I am of the mind that a low signal boosted radically by a limiter to achieve the required loudness level is not the same aurally and sonically as a 0db signal boosted less by the same limiter to achieve the same loudness level.

But normalizing also changes the sound of the wave as it's pretty much digitally bringing the loudest peak to 0db. I'll have to find this info but there was an article about this that I read when I was starting in the domain.

For certain no-one who masters in this way can talk with any confidence about the degree of db boost that has been applied, because they don't know what their starting point is.

I don't see how this is important? the point is what you reach, now what you did...

[aj113]

...

I'm curious about why it is important to see it?

Because it gives me more control. I know what commercial wavs look like. I know what quiet wavs look like. I'm not saying that making your waveform look like the big boys' waveform is going to suddenly make it sound like the big boys, but it's definitely not going to sound like the big boys if it doesn't look like the big boys. Hope that makes sense.

The point of the limiter is to bring it to 0db (or in my case I put -0.1db just to be sure in case the limiter misses something)
That's how you know your final mix hit 0db is with the limiter.

The point of the limiter is take low transients closer to 0db than they curently are. The point of normalising is to take your peak signal to 0db and adjust the rest of the waveform on a pro rata basis. Hence my point: If you are relying on a limiter to take very low transients up to the required level rather than simply taking the level up first with a fader (or by normalising) before limiting, then surely the sound is going to suffer in some way?

But normalizing also changes the sound of the wave as it's pretty much digitally bringing the loudest peak to 0db.

And how does that change the sound? Ever since I started recording digitally I've had it preached to me that getting a good recording level is not necessary when you are recording digitally as there is no noise from digital recording, and the signal can be taken up as high as you like after the fact (before clipping of course) without affecting the quality of the recording.

[garageband]

(or in my case I put -0.1db just to be sure in case the limiter misses something)

Or even -.2. You don't want to add dither and have that push the program beyond 0dBfs.

[acidfrost]

But normalizing also changes the sound of the wave as it's pretty much digitally bringing the loudest peak to 0db.

And how does that change the sound? Ever since I started recording digitally I've had it preached to me that getting a good recording level is not necessary when you are recording digitally as there is no noise from digital recording, and the signal can be taken up as high as you like after the fact (before clipping of course) without affecting the quality of the recording.

This is probably BS from my part as it's something I read a while ago. I just have a fear of normalizing haha!

then surely the sound is going to suffer in some way?

This I don't understand how it's suffering. Yes, most plugins and hardware change the sound but to what point? I don't think normalizing is better than just bringing up with a limiter. Could be all this visual thing of yours :P lol, but why do you do so?

(By the way, I'm not trying to argue or diminish your points, I'm just trying to understand your way of doing )

[shackman]

And how does that change the sound? Ever since I started recording digitally I've had it preached to me that getting a good recording level is not necessary when you are recording digitally as there is no noise from digital recording, and the signal can be taken up as high as you like after the fact (before clipping of course) without affecting the quality of the recording.

That's right. Digital recording creates no noise. Tape DID cuase it's own noise (Ie, play a silent stretch of tape and there was a tiny level of noise intrinsically part of the sound.

BUT: every recording you make througha a mic or a pick up has it's own noise placed INTO the digital realm.

So. If you raise ALL the elevls "pro rata" by normalising, you also raise the noise inherent in "live recording".

It's like the noise is broguht into digital by the cables, nmics, pickups etc and printed in the digital realm.

Therefore, when you normalise, you change the recording!

You could find the noise floor and gate it, but that's messy.

I have no answers, just observations - if you're still talking to me!!

[aj113]

.... If you raise ALL the elevls "pro rata" by normalising, you also raise the noise inherent in "live recording".

That may be true, but I'm talking about normalising in place of limiting. In other words, normalise to 0db using peak levels vs "normalising" to peak levels using a limiter. Is it the same thing? I don't know enough to call it, but my head is telling me it's not the same thing, and that the limiter is going to add potential unnecessary artefacts, i.e. mud.

[bozmillar]

I am of the mind that a low signal boosted radically by a limiter to achieve the required loudness level is not the same aurally and sonically as a 0db signal boosted less by the same limiter to achieve the same loudness level.

I don't think that this is at all true, at least not with any plugin limiters. And this is extremely easy to test. Take a song (unmastered) that has a peak value of, say, -20dB and put a limiter on it with your threshold at -28dB or whatever.

then duplicate the original track, normalize it to 0db and put the same limiter with a threshold of -8dB and see if they null.

And the second issue I have, is who says that 0dB is optimal for the limiter? If you're working with hardware, 0dBFS is rarely it's optimal level. If you are working with software, it doesn't matter.

[garww]

As far as I know, normalizing is nothing but a log expander. I would experiment with what you got. For me, I am more likely to use it first getting my tracks(90pct) at a workable level from the start. And, I may use it again for final(90-95pct). I haven't played with 0db. A fair number of the songs I get off Amazon and some other places are over 0db on my meters

[venuestudio]

I have read for years to stay away from normalizing. That comes from mastering engineers. It is a lot like taking your track, sending it to someone else to compress and then getting back the results. You are out of the loop. Shackman is correct, the noise level will be raised too. Compressors don't create mud unless you drive them to that point. Compressors can be very, very clean. Limiters too. It just depends on how they are used.

I suppose there are times normalization is fine but for me I would start by working the individual tracks to get to the loudness you want (assuming that is what you are shooting for).

[rook2c4]

I have read for years to stay away from normalizing. That comes from mastering engineers. It is a lot like taking your track, sending it to someone else to compress and then getting back the results.

Strict normalization (which doesn't include limiting or compression, only overall level adjustment) shouldn't do anything to compromise the sound. Yes, noise levels are raised, but so is everything else.

For example, take an audio file. reduce the volume by 1 dB. Then take that new file and increase the volume by 1 dB. Any difference between the original audio and the final audio - if there is one at all - will be so miniscule, you won't be able to hear it. Whatever difference there is, will be tiny processing artifacts.

I think mastering engineers that say stay away from noramalizing do so because it usually adds an extra step for them: reducing the level enough so that they can apply additional processing. Normalizing a file before mastering is basically a pointless activity, but not really a harmful one.

[Stan_Halen]

Another interesting viewpoint AJ! I like how you bring a fresh perspective on the "how and why" of the processes we tend to take for granted.

I have never heard of anyone Normalizing a song .wav file prior to mastering. Then again, so far I don't think we have been able to give a strong argument why you shouldn't do it. I don't know if I can add anything to that, though it would seem there is some reason that multitudes of engineers and mixers use limiters to achieve the result and don't use Normalizing (not that I have heard of anyway).

Normalizing is fine, and recommended, for certain tasks: such as boosting a voice recording done in a quiet vocal booth, to match a standard that will be used in a finished product (i.e. Audiobooks, commercials).

Normalizing brings up both the signal (average or peak) and the noise floor. So you get more signal, but you also get more noise. The signal-to-noise ratio doesn't change, you're just making everything louder.

Limiters, and compressors, make the loud parts softer and the soft parts louder. You might conclude that this isn't that much different than Normalizing (making the noise floor louder), but there probably are significant differences in signal processing. The process of compressing changes the perception of the loudness in the signal, and perhaps the noise floor as well. Normalizing does no such changing, it just boosts.

I can understand that you like the visual aspect of the wave-form. It gives you more of a sense of control than handing that control over to a software mathematical formula. However, 99% of us (wild guess) believe in the Compressor/Limiter God that brings everlasting life in the Kingdom of Loudness, and so we worship it eternally.

The one thing that really shocks me is that you Normalize to 0 dB. But then you're using a limiter to reduce that so you don't clip the signal (?). Why do you need a limiter at all then? I wouldn't Normalize higher than -0.3 dB or so. At that point I don't know what a limiter is even doing for you though ...

[garww]

You can do a sine wave and watch the amplitude change as you play with levels. You can normalize and see how it compares, but amplitude on a simple signal should be pretty simple math. If the signal contains essential room detail then you need to be sure and maintain that aspect with other tracks or it can screw you up. For me, it's usually a weak DI signal. I wouldn't assume all the software vendors use the same method and some probably just suck. Amplitute should be one of the least signal degrading math functions - what is that little make-up knob off to the side of a vst compressor doing ?

I'm lookking at the Help file for a XP era Magix product and they explain the different functions available along with a few cautions. Under the option to "norm" album mastering, they say they bring in their MultiMax limiter automatically when needed. They manuals are more detailed, of course. RTFM, but test it to see if it holds up.

EDIT; I forgot to say that, as with anything else, you can certainly find your way into a bottomless pit if you're without a plan from the start.

[Stan_Halen]

Limiters, and compressors, make the loud parts softer and the soft parts louder. You might conclude that this isn't that much different than Normalizing (making the noise floor louder), but there probably are significant differences in signal processing. The process of compressing changes the perception of the loudness in the signal, and perhaps the noise floor as well. Normalizing does no such changing, it just boosts.

I found an article that explains much more eloquently what I was hoping to get across:

Tech Tips: To normalize or not to normalize during the mastering stage. | Sweetwater.com

[bobbybovine]

Because it gives me more control. I know what commercial wavs look like. I know what quiet wavs look like. I'm not saying that making your waveform look like the big boys' waveform is going to suddenly make it sound like the big boys, but it's definitely not going to sound like the big boys if it doesn't look like the big boys. Hope that makes sense.

This has already been said but use your ears more than your eyes. Your eyes can trick you into thinking you are hearing something you are not. If it sounds as loud as your reference, and the TT meter or other meter is telling you it is the appropriate RMS the what do you care what the waveform looks like. I guarantee if you get to -6db rms or lower you will have the brick you are looking for.

The point of the limiter is take low transients closer to 0db than they curently are. The point of normalising is to take your peak signal to 0db and adjust the rest of the waveform on a pro rata basis. Hence my point: If you are relying on a limiter to take very low transients up to the required level rather than simply taking the level up first with a fader (or by normalising) before limiting, then surely the sound is going to suffer in some way?

And how does that change the sound? Ever since I started recording digitally I've had it preached to me that getting a good recording level is not necessary when you are recording digitally as there is no noise from digital recording, and the signal can be taken up as high as you like after the fact (before clipping of course) without affecting the quality of the recording.

In a 24 bit world and beyond (32 bit floating point) noise is not a concern, you have tons of headroom. You need to think of this from a purely digital standpoint since you are talking doing this in your DAW. All normalizing is doing is adding another calculation before the limiter. Even a .1 db level adjustment is a calculation. So the goal with digital mastering is to get to the desired result with as few calculations of the data as possible. Calculations bring rounding into the equation or truncation which can drastically alter your sound in the 16 bit domain. But in 24 or 32 bit the headroom is so high that this is usually not an issue, however there is something to be said for best practice. If you can reduce the amount of steps to the final result then you should do it. Now in addition to this, and contrary to what I just wrote, I usually use 3 limiters before the final result. I find using one limiter to make up all the rms level never sounds as good as bumping it up in stages with 2 or 3 limiters. Each one with the peak level set closer to 0 than the rest. I always have my last stage at -.2 db to make sure there is absolutely no clipping. So I am not saying don't normalize but I think it is an unnecessary step.

[garww]

I found an article that explains much more eloquently what I was hoping to get across:

Tech Tips: To normalize or not to normalize during the mastering stage. | Sweetwater.com

Call you S* Sales Engineer !! Good advice : ) Ha !

[bozmillar]

Normalizing a file before mastering is basically a pointless activity, but not really a harmful one.

This pretty much sums it up.

[garww]

"So I am not saying don't normalize but I think it is an unnecessary step".


I'm currently doing my groups to a mixer or two for normilization in my fake mastering chain to tape

[bobbybovine]

"So I am not saying don't normalize but I think it is an unnecessary step".


I'm currently doing my groups to a mixer or two for normilization in my fake mastering chain to tape

This is where normalizing makes more sense, to make sure you get the hottest signal since your noise floor is higher.

[garww]

My normalization may include dynamic range expansion, too

[bozmillar]

It always amazes me that the most simple possible process you can do to audio (normalization, aka volume change) can carry with it so much mystery. Normalization does nothing more than turn your signal up or down to have a specified peak value. If the thought process goes any deeper than "Do I need to change the volume on this?" then you are thinking about it too much.

[brandondrury]

1. You can't see the resultant waveform either before or after the limiter - both are important to me.

Interesting. No, I can't say I really ever care what the waveform looks like. There are a few pieces of information I do need to know beyond what my ears tell me and most of that can be done with meters.

2. You can't be sure that the signal before the limiter is normalised to 0db.

No, and I don't care necessarily. All my brickwall limiters have gain level so I'm good to go.

I am of the mind that a low signal boosted radically by a limiter to achieve the required loudness level is not the same aurally and sonically as a 0db signal boosted less by the same limiter to achieve the same loudness level.

This is an interesting hypothesis, but it's only a hypothesis. What's your evidence? While I'm sure someone has made a limiter with specific non-linearities in it, this would be a lot of work for them to get right. Either way, start a new thread with you test mp3s. I'd love to hear it.

I do recommend studying how floating point math works. You'll quickly realize how levels aren't that big of issue within the digital domain until we do the final render.

For certain no-one who masters in this way can talk with any confidence about the degree of db boost that has been applied, because they don't know what their starting point is.

You do realize that the brickwall limiter bypass button is just one click away, right? I still need to be convinced that my 2bus needs to be peak normalized before hitting the limiter.

I'm also curious why you believe that the digital process of peak normalization is transparent, but the gain in front of a brickwall limiter is not transparent.

Because it gives me more control. I know what commercial wavs look like. I know what quiet wavs look like. I'm not saying that making your waveform look like the big boys' waveform is going to suddenly make it sound like the big boys, but it's definitely not going to sound like the big boys if it doesn't look like the big boys. Hope that makes sense.

A person should use all their senses with this gig. They just need to trust their ears a hell of a lot more than others. It never hurts to take a look at your stuff verses the big boys, but that is usually just a clue that you'll have to solve with your ears.

I have read for years to stay away from normalizing.

Peak normalization is nothing more than gain and a calculator. I've not used it in a decade, but some people want to find their peak, do the math to get that peak to 0dbFS and the peak normalization function will do that. Of course, if you have to peak normalize you aren't anywhere near modern RMS loudness. The brickwall limiter makes peak normalizing obsolete in my shop.

RMS normalization is a non-musical compressor that will destroy any song/mix you put into it. NEVER EVER EVER do it unless you just want to ruin what you've got.

[garww]

It always amazes me that the most simple possible process you can do to audio (normalization, aka volume change) can carry with it so much mystery. Normalization does nothing more than turn your signal up or down to have a specified peak value. If the thought process goes any deeper than "Do I need to change the volume on this?" then you are thinking about it too much.

That's it in a nutshell, but if you take random audio samples from youtube and "master" with a album in mind it's gets a bit more complicated. Any of that kind of Mastering is a good execise.

[bozmillar]

That's it in a nutshell, but if you take random audio samples from youtube and "master" with a album in mind it's gets a bit more complicated. Any of that kind of Mastering is a good execise.

yeah, but normalizing isn't mastering. Not even close. In the same way that brushing my hair isn't the same thing as driving to work. Normalizing is a possible step that some people might use before mastering.

[garww]

yeah, but normalizing isn't mastering. Not even close. In the same way that brushing my hair isn't the same thing as driving to work. Normalizing is a possible step that some people might use before mastering.

The thread opens with;
"Another point which has got me thinking recently:

My current method of mastering is to import the rendered stereo track into Soundforge and then normalise to 0db. From there I have a known quantified platform (i.e. a wav normalised to 0db) from which to operate my limiter.