Montage loudness / meta-normalizer questions

Wavelab 7 MetaNorm Settings.PNG
Wavelab 7 MetaNorm Result.PNG

If you bring all of the clip volumes down 1db (with the volume envelopes) and re-run the Meta Norm, it “starts” to do volume matching (but only partially, only what it has headroom for). If you bring the clips all down 2db, it does more volume matching. Down 3db, even more. Until it finally has room to entirely volume match under zero. This just doesn’t seem right, because the ability to do any volume matching at all is dependent on the original levels of the clips, so it’s probably going to be wrong in almost all cases of normal / loud clips.

WaveLab takes into account the volume envelope, hence what you say is a normal situation.

Wavelab 7 took care of this by reducing compared to the loudest clip (not raising) if a zero ceiling was set. It would put the entire montage peak at zero, and lower all clips relative to the loudest clip. (see Wavelab 7 png’s in the following post).

WaveLab 8 changes the overal gain by changing the (new) master gain. WaveLab 7 was changing the gain of each individual clip. Sonically speaking, this is exactly the same result if there is no track effect. “Visually”, there is the difference you describe.
If there are track effects including compressor-like plugins, WaveLab 8 will give some more exact result, because the correction happens after the effects.

Thanks PG. But aren’t the “Track” plugins being operated in the “Montage” section of the Meta Normalizer, after the “Clips” section, so it shouldn’t matter whether the Clips are scaled up or down? And if you’re using Track plugins then you should know to engage the “Montage” section?

My point is that the “Clips” section is not operating as it should (it should be volume matching under all circumstances), using only the Clips settings that the program allows me to use as seen in my settings pic. If Clips were scaled down it would always operate as it should I would think. And the track plugins would be normalized as they should if you engage the “Montage” section for that, as in the Factory Default preset. But that comes after the Clips are scaled up or down.

The other thing I noticed is that Track plugins are being hit with unusually hot incoming levels if Clips are being raised 3 or 4 db above 0 as they are in the Factory Default preset on already loud material. I can see this in track plugins that have Input Meters going seriously in the red at input. Maybe it doesn’t matter, everything has enough headroom, but those very hot levels are what’s hitting the input of the Track plugins, and it seems very strange. And the input level to a Track compressor plugin is changing after the Meta Normalizer analyzes the entire chain, which may or may not be wanted. Of course scaling clips down would be going the other way as far as the input to the track compressor, so I don’t know if there’s an answer or methodology for doing that sort of thing.

It just seems to me that the Clips section should be scaling down rather than up, because that section is not doing what it should be doing if used independently and a zero ceiling specified. With my settings it won’t do any volume matching at all on Clips that are already near zero. And it will only do partial volume matching as you incrementally lower the Clip levels from there. In many cases you would need clips to all be down 6 db in order to do the level matching if there was that much difference between the clips. And even on this finished commercial record, levels are being normalized by 4 db, so I don’t think that would be unusual.

I know Meta Normalizing an already mastered finished record is an extreme case, but I don’t think the example is unheard of. I think anyone can see this for themselves by loading in most any pop / rock commercial release and running it through the Meta Normalizer.

The other good thing about the Wavelab 7 method was that after rms volume matching the clips it would also raise everything to have the montage peak at 0 or -0.1 if that was the ceiling you set. I don’t see any way to do that in the Wavelab 8 clips section.

Thanks PG. But aren’t the “Track” plugins being operated in the “Montage” section of the Meta Normalizer, after the “Clips” section, so it shouldn’t matter whether the Clips are scaled up or down?

Yes. But in WaveLab 7, contrary to WaveLab 8, if the level is changed “not linearly” by a track plugin, this causes a small computation error, because the level correction (determined by the analysis at the montage output) is applied on the clips, hence prior the track plugins. In WaveLab 8, the correction is applied post the track section, hence the computation is exact.
I know it’s maybe a bit difficult to understand…

The other good thing about the Wavelab 7 method was that after rms volume matching the clips it would also raise everything to have the montage peak at 0 or -0.1 if that was the ceiling you set. I don’t see any way to do that in the Wavelab 8 clips section.

The montage section in WaveLab 8 does exactly the same, but without changing the gain at the clip level.

Philippe

I understand what it’s doing, but in the example I gave, if the Factory Default Preset is used on a normal loud level commercial release, all of that “up” clip scaling is being fed to the input of the Track plugin. I’m seeing +6 peaks from some clips on the Input meter of an Oxford Inflator placed on the track. Can the plugin itself handle that?

Also, would you say the Clips section is operating correctly if (when used independently) it can’t volume match average to loud level clips using the settings I pictured? Honestly that seems like a pretty obvious bug, and the clips should be scaled down, not only up.

Anybody can see all of these things loading in a normal loud level commercial release. I realize that’s a bit of an extreme example but it’s not unheard of.

PG, will this be fixed in an update? Clip section independent usage, and Factory Default Preset usage without Track Plugin input overload?

If you read my previous answers, you can see that there is no bug in the current implementation, nothing to fix.
The only thing I agreed, is to have an easier way to peak normalize the clips (something which is, indirectly, already possible). This new option will happen in the future, but not in a bug fix version (ie. not in 8.0.3).

PG, I’m sorry, but I really don’t understand why this is not going to be changed. I’ve read your posts, I understand how it works, but it’s not correct, and it’s unnecessarily risky with the track plugin overload problem. That is just simply asking for trouble. Clips are being raised, but no one would ever manually push clip levels up to what the meta normalizer is doing. While I understand levels are normally offset in the Factory Default Preset, as soon as you throw a Track plugin in there you’re just asking for the Input of that plugin to overload because that point in the chain is not addressed at all in the MN gain staging.

Also, the Clips section on its own (as in pic below) simply doesn’t work to level match tracks from a commercial CD.

There are 3 problems, and they’re all due to the fact that clip levels are only raised in WL8, not lowered, as in Wavelab 7. That causes all 3 problems, including the OP’s problem with the visual waveforms.

  1. Clips appear as massive unintelligible waveform blocks because clip levels are raised to absurd levels by the meta normalizer (the OP said some of his by 9 db).

  2. These absurd levels are fed to the Input of the first Track plugin, overloading the input. The answer is not to lower the input level on the plugin (if it has an Input level adjustment), because that would require running it first to see how much, lowering and then re-running the normalizer. The answer is to match clip levels by lowering the clip levels relative to the loudest clip in the first place, like Wavelab 7 did. Also some plugins don’t have Input level adjust so are going to be overloaded no matter what.

  3. The Clips section in the MN if used independently (Montage and Master Section unchecked) doesn’t level-match tracks from a commercial CD, as it should. It only changes the levels of the clips by a tiny fraction of what it needs to level match, because it can’t turn the Clips down. That is definitely a bug. (settings pic below).

Can someone at Steinberg please just try this by loading a few commercial pop or rock CD albums, and running the Factory Default Preset with a plugin on the Track (like the Steinberg EQ or Limiter). The plugin Input will invariably be overloaded. This is not any kind of normal gain staging.

And also run the Clips section independently as in my picture. That simply doesn’t work as it should, and is a bug. (It doesn’t level-match the tracks because it can’t turn the Clips down).

Can you please reconsider this because it’s simply not right. I’ve shown this to other people, and they agree. And all 3 problems would be solved by simply allowing Clips to be lowered in level like they were in Wavelab 7.
Wavelab 8 Settings 2.PNG

There are several ways to use the meta-normalizer, and this also depend on your music. In your case, if you set “match loudest clip” and the loudest clips is already heavily compressed (high loudness), then yes, some of your other clips are going to get a high gain, something that is maybe not proper for them, because their content was not as much compressed.
In that case, you could take another loudness reference (not the “loudest clip”). Or you could exclude the loudest clip(s) from the process.

Understand that when you start with audio that is heavily compressed, you start from “un-natural recordings” (this is not pejorative here). Hence if you want to reach loudness balance between all clips, they must all have been compressed somehow in the same way. The meta-normalizer is not a substitute to compressors/expanders.

Concerning the track section, I recall that the audio bus is 32 bit float, hence there is never clipping until the very output mix of the audio montage. That being said, with the option “Limit peaks”, you have control on that, to prevent > 0 dB peaks at the clip level.

BTW, concerning the pictures you showed earlier between WaveLab 7 and 8, I note that there is a sensible difference: in WaveLab 7 you try to match the loudest 2 seconds section (because “Global” is unchecked), while in WaveLab 8, you try to match the average song loudness. This being said, the underlying algorithm is in any case not the same, but this setting difference could be important. The closest WaveLab 8 setting would be “Reference: Maximum short term loudness”)

I think Bob99’s points are well taken, and this is indeed a LOF relative to WL7 as he’s explained. I welcome the new algos, but miss the old approach. I can understand that this can’t make 8.03 without disrupting actual bug fixes, as this change would me a new feature/implementation in WL8 (even though it existed in WL7).

I take your point about the 32 bit bus, PG, but as an old timer with analog gear and outboard digital I patch in find that proper gain-staging still matters. Even many plug-ins sound better when not over-loaded or driven too hard. While we have vast headroom in 32bit float, the reality is our signals exist in a space smaller than 24 bits in the analog and outboard realm; sure 32 bits scale nicely to 24 and what’s lost is error or irrelevant data, no argument. But there’s value and elegance in clean, intelligent gain staging (if only not having to re-think things when you patch something sensitive or level dependent in the chain).

At any rate, for many users this is a matter of principle, and in mastering this approach is pretty normal: cut first, boost later.

I will study the possibility to have a mode similar to what was in WaveLab 7.

Hi,

Is the “old” Meta Normalizer available again in a new version of WL8?
We’ve switched back to WL7 because of the lack of this feature in WL8.
Working with workarounds to do this is not an option, sorry.

Thanks!
Philippe

No, it is not. But the idea is not forgotten.
Unlike you, I think workarounds to do this are an option…

Thanks for keeping this in mind, PG.
I could use workarounds if I was the only user.
However, we use Wavelab in an institutional environment with external freelance users as well as student workshops, etc.
Workarounds can’t be used to teach students, or explain to freelancers, how to do a common task in a “strange way”…
best
Philippe