Normalizing Compilation CD w/ Acid or SoundForge

[Puño de Dolares]

Good afternoon. I'm in the process of making a soundtrack CD and I have several tracks with volume levels a tad too high or a little low ... I'm trying to find how to normalize all the tracks at once (using Acid/SForge/or Vegas), but couldn't find it in the manuals.

 

If anyone knows, I'd greatly appreciate it if you'd pass on the knowledge. Thanks.

[JimmieWannaB]

More knowledgeable members can probably go through the tweaking details with you, but I typically just use the Normalize defaults in Sound Forge.

- Process

- Normalize

- Average RMS

- Normalize RMS to -16 db (music)

- If clipping occurs: Normalize peak value to 0 db

[DWBass]

You should be able to normalize in any cd burning program/software. I use Nero to normalize all tracks before burning.

[doug osborne]

Most mastering engineers recommend no normalization.

 

With 24-bit recording, we have enough headroom that resolution will not suffer if one song is a few dB higher or lower than another in peak values. But, simply raising all the bits in each song so it is at your chosen ceiling will not make each song sound, on average, appropriately loud.

 

Normalization might make sense to make a number of small clips recorded at somewhat different levels more like each other so they can be bounced and remixed, but even then, the clips might not sound the way you want or expect.

[DirtyRubberDuck]

With 24-bit recording, we have enough headroom that resolution will not suffer if one song is a few dB higher or lower than another in peak values. But, simply raising all the bits in each song so it is at your chosen ceiling will not make each song sound, on average, appropriately loud.

So you're saying that Puño de Dolares can't or shouldn't normalize songs to get the volume level the same? With no disrespect to your mastering abilities, I think you may have lost sight of the forest from the trees.

 

DRD

[Rog]

I wouldn't normalise to 0dB. Even though that means that the whole track is raised until the sample with the most amplitude reaches 0dB, the DA convertor reconstructing the waveform can actually exceed 0dB and lead to clipping.

 

I'd use your ear and the average RMS of each song or average RMS of the loudest part of each song as a guide.

[doug osborne]

Originally posted by DirtyRubberDuck:

So you're saying that Puño de Dolares can't or shouldn't normalize songs to get the volume level the same? With no disrespect to your mastering abilities, I think you may have lost sight of the forest from the trees.

 

DRD

Yes, that's what I'm saying.

 

Normalization takes the highest-amplitude bit in an audio clip and raises the whole clip so this it brought up to your theoretical limit. This may have been necessary to preserve resolution in 16-bit recording, but it is generally considered unimportant with 24-bit recording, and can even be detrimental, as Rog said.

 

I'm not saying that you don't want all tracks to be at their proper perceived level, but the automatic "normalization" command might not be the way to go. Individual Volume Envelopes per song in SF might be more appropriate.

[DirtyRubberDuck]

Yes, that's what I'm saying.

 

Normalization takes the highest-amplitude bit in an audio clip and raises the whole clip so this it brought up to your theoretical limit. This may have been necessary to preserve resolution in 16-bit recording, but it is generally considered unimportant with 24-bit recording, and can even be detrimental, as Rog said.

 

I'm not saying that you don't want all tracks to be at their proper perceived level, but the automatic "normalization" command might not be the way to go. Individual Volume Envelopes per song in SF might be more appropriate.

I still think normalization is the best way to go. It does the same thing as a volume envelope, but it does it automatically. You set the values for the normalization function in one song, and use the same setting for all the songs. (I agree to use ave. RMS and not peak) Sound Forge has a batch processing function to do all the songs at once. You have to keep the normalization level low enough to avoid clipping, but you have to watch that with volume envelopes too. And you don't have to set the levels for each song individually.

 

As for the resolution, ASSUMING that our buddy Puño is using 24 bit recording, he shouldn't have to normalize to "preserve resolution", but then, that wasn't really the question, was it?

 

DRD

[doug osborne]

I guess our friend Puño is going to have to try both ways and see which works best for him, isn't he?

 

We're no help here, are we?

[Anderton]

The problem with normalizing is that it's not a "one size fits all" solution. With material that has the same general dynamic range and AVERAGE (RMS) level, normalization can work. But suppose you have a real rocker, lots of distortion...it will have a really high average level. Now you have a ballad with a lot of percussion and acoustic guitar. The peak levels may be quite high, but the average lower will be much lower. Although the rocker should probably sound louder than the ballad, if normalized, the difference could be HUGE.

 

In a case like this, you might want to bring the level down on the rocker a dB or two, and add some multiband compression to the ballad so it comes up a bit in level but doesn't lose too many dynamics.

 

What I suggest is try normalizing each tune if you want, put them into a playlist, then listen to the entire collection. If a song sounds significantly louder than the others, bring down its overall level in the playlist.

 

Just remember that you don't want to apply too many DSP processes, like change level, normalize, change level again, etc. because each time you do so may subtly degrade the sound. It's best to apply any needed DSP processes only once, if possible.

 

I hope this helps demystify things a bit!

[Lee Tyler]

Puno. How about stacking up all the tracks in a multitrack audio editor and make judicicious changes in volume (normalizing may be a no-no..see link below) whilst soloing individual tracks in many different spots, and use something that is called *judgement. *(a developed skill that some excell in better than others) Works for me and all my subsequently treated tracks will sound dead-on-balls, balanced, pro-quality without loss of quality/clarity. ---Lee

 

S.O.S. discussion on normalizing. (brew some coffee ot tea..this will be a long night)

[Lee Tyler]

Craig, didn't even read your post! Guess we are getting at the same thing. ---Lee

[Anderton]

<>

 

Well as they say, "Great minds think alike"...or maybe it was "lesser minds run in the same gutter." Or something like that!

 

With dance mixes, I do normalize my tracks during assembly so I have a common point of departure, then tweak levels prior to dithering and running off a test CD. Most of the time, though, levels don't need to be tweaked much, if at all.

 

But the only reason I can get away with this is because the tunes are all mixed under the same circumstances by the same person using the same gear and listening with the same ears to a more or less calibrated listening space, so the mixes naturally tend to be the same level anyway.

[M Peasley]

My habit is, before I normalize a track, I examine the waveform in Sound Forge for big spikes in volume. Since normalizing only raises the level until the loudest spike reaches the set -db limit, a loud spike can greatly hinder the normalization. If there is a really whopper of a spike, it might even be louder than your set -db normalization parameter, in which case normalization will only REDUCE the overall volume.

 

Sometimes I compress the track a bit to bring the high spikes and average levels closer together. Sometimes I zero in on the specific spikes and simply reduce their volume using Sound Forge.

 

That way, when I normalize, I get a lot more average level increase.

 

I'm not sure I'd ever normalize multiple tracks at a time. Maybe mastering engineers do this, but not me.

 

M Peasley

[Anderton]

I don't know of any mastering engineers who normalize multiple tracks at a time, but that of course doesn't mean no one does it.

 

As to those peaks, I try to get rid of them while mixing so they're not an issue in mastering. I do this by checking the output meters and wherever there's a significant peak, I find out which tracks are combining to cause the problem, and reduce their level for a few milliseconds.

[Lee Tyler]

I find out which tracks are combining to cause the problem, and reduce their level for a few milliseconds.

Ah! A fellow millisecond-micro surgeon. I often find myself diving into that sort of "musical" engineering for volume and many other sound-shaping parameters... and not only does it make a big difference in a track in toto, it is kinda fun doing such procedures as well. ---Lee Em Dee.

[DirtyRubberDuck]

I beleive that you can get around the "loudness" vs. peak level thing by using ave RMS setting in Sound Forge as suggested above by JimmieWannaB. You DO still have to check for stray peaks, which could screw up the whole process. If Sound Forge detects that the normalization process via RMS will result in a clip, it can either revert to "normalize peak value to zero" which, I believe, renormalizes the whole file using that particular peak, normalized to zero as the maximum, thereby screwing up the whole RMS thing. The other option is to apply compression, which also affects the entire sound file. So you do need to check for those stray peaks.

 

DRD

[miroslav]

Originally posted by Puño de Dolares:

...I have several tracks with volume levels a tad too high or a little low ...

Maybe I'm missing something here from all you "normalizing addicts" ...

 

...but, isn't normalizing intended to "level" out each song.

That is...it will bring up or lower the peaks/dips so that the level of the song is kinda' even from start to finish...right guys?

 

But...Puño de Dolares is asking about the levels of all the songs relative to each other...

...that's way I'm reading his post.

 

My solution would not be to normalize...but just adjust the overall level of each song relative to the other songs.

That will still preserve the dynamics within each song...whereas normalization tends to kinda' even out all the dynamics...depending on hard you go for it...sorta like compression does.

 

You have to pick one song as your reference level songand then adjust the overall level of the others to it.

Of courseyou have to also consider the level of your reference songcuz if it is too close to digital 0

...then you may be painting yourself into a corner if you want one of your other songs to be louder.

 

Anywaynormalization is a cheap way outIMO.

Its likeI screwed up the mixso Ill just normalize it into submission

[doug osborne]

Originally posted by miroslav:

...

 

...but, isn't normalizing intended to "level" out each song.

That is...it will bring up or lower the peaks/dips so that the level of the song is kinda' even from start to finish...right guys?

 

...

Miroslav, I think your conclusion is correct, but there is still confusion on the concept of normalization.

 

The normalization command will typically find the loudest peak in an audio clip, and re-calculate the song so that that peak is moved up to the theoretical or chosen maximum, possibly -0.02 dB from 0.0 dBfs. Compression will squash this peak, normalization won't.

 

Imagine a loud rock song with one loud peak, and a quiet ballad with no such peaks (things usually come out the other way around...). The rock song would be normalized to that peak, but the ballad would be brought up to sound much louder. This is probably not the effect you want.

 

As has been explained, normalization was invented to take advantage of all the bits available in 16-or-fewer-bit recording. This is typically not important with 24-bit recording, in fact, the fewer the calculations, the better. Compression or limiting at this stage may be appropriate depending on the content, but it still isn't the same as normalization.

 

If the goal is to make all songs on a compilation to be at a similar percieved loudness level, neither peak nor RMS normalization can be trusted, but your ears can be trusted. Again, at this mastering or pre-mastering stage, my recommendation is to use a volume envelope or automation per song (or per song-section, even) to bring this song or section up or down to meet the desired level, but there are many ways to do this without a one-size-fits-all normalization and without downward compression.

[Anderton]

<>

 

If you're using a digital audio editor with a playlist, you can usually adjust the level of each playlist entry.

[miroslav]

Originally posted by Doug Osborne:

Miroslav, I think your conclusion is correct, but there is still confusion on the concept of normalization.

I guess my conclusion stems from all the options that are given in the Sound Forge - Normalization window.

 

It allows you to (w/RMS setting) adjust Attack Time, Release Time...

..and you can also mess with the Equal Loudness Contour....and apply Dynamic Compressionetc.

 

So...there's a LOT more that can go on there than just a straightforward level increase/decrease.

 

I guess that is why I find "normalization" a rather questionable process...whereas just lowering or raising the overall level is fine...

and it looks like in SF you can use that minimalist approach too, when applying Normalizationthough I prefer to just raise or lower a fader.

 

I'm still only using my DAW for very basic cut & paste edits and track sub mixes...

...everything else happens outside the DAW when I go back to analog for the final mix/processing.