Normalise Regions for gain staging not working as expected

[jmob]

5 minutes ago, JoshJ said:

Supposedly the unwanted artefacts of aliasing are cumulative. So using my earls alone presumably would be a challenge. Having a target reference would help and to reduce decision fatigue. The Normalise Region function is helpful. Hornet make a couple of plugins that normalise the signal to a target also.

I'll tell you what I do. I select all of the tracks and pull them done until the drums (whatever it is that is the focus of the track) hit -12 db. Then, I mix from there. You end up with a ton of starting headroom. It gives you a lot of room to make instruments louder or softer etc. Once I started doing this my mixes became a lot cleaner and clearer.

[triplets]

10 minutes ago, JoshJ said:

So using my earls alone presumably would be a challenge.

The best sounding mixes that we praise still today from 70s, 80s and 90s, where all based on ears. None of this "let the computer or A.I. decide" that's taking the mixing world by storm today.

[dubsak]

6 hours ago, triplets said:

You're overthinking this way too much. If you're keen on normalizing, then use the default average loudness of -23 LUFS for individual regions in Logic, that way any peak will still be within analog parameters. Then start using your ears and don't mix with your eyes only.

Like Triplets says, Normalizing at -23 LUFS will work for a lot of things (think Bass, Guitar) but won't work well for others (Snares, Kick Drums or anything that's all about the transient). It's kind of why Normalize Region Gain doesn't work so well for music (as the source is all over the map unless you're dealing with mastered, complete songs or other material) so there's no definitive answer. You likely want to just watch your peak meters in the mixer and just see that peaks are in the yellow and the meter is mostly always in the green around -18 dBFS (again, it depends on the source). When people are saying things like "plugins are optimized or designed for 0 dB on a VU meter", that's what they mean. The best way to accomplish this (after the fact assuming you didn't gain stage into Logic to your satisfaction), is to use the region gain tool to raise or lower your gain per track by sight using those meters or the MultiMeter as Matthew recommended. Even with absolutely perfect gain staging, once you start using saturation, depending on the amount, you're going to increase the harmonics upwards quite a bit past 20kHz and there will be aliasing without oversampling. So…long story short, it all always depends. Kind of hard to have a definitive answer here.

[jmob]

15 minutes ago, triplets said:

The best sounding mixes that we praise still today from 70s, 80s and 90s, were all based on ears. None of this "let the computer or A.I. decide" that's taking the mixing world by storm today.

What he said. Mixing is always a challenge. 

[matthewbarnhart]

22 minutes ago, triplets said:

The best sounding mixes that we praise still today from 70s, 80s and 90s, where all based on ears. None of this "let the computer or A.I. decide" that's taking the mixing world by storm today.

True, but those records were also made by well-trained engineers who understood mic selection, mic placement, arrangement, analog gain-staging, tape recorder calibration, acoustics, etc... all recording a lot of really excellent musicians.

Sgt. Pepper's was indeed made on a 4-track, but it sure wasn't made on a PortaStudio!

[polanoid]

38 minutes ago, JoshJ said:

Supposedly the unwanted artefacts of aliasing are cumulative.

What would gain staging have to do with aliasing?

[jmob]

7 minutes ago, matthewbarnhart said:

True, but those records were also made by well-trained engineers who understood mic selection, mic placement, arrangement, analog gain-staging, tape recorder calibration, acoustics, etc... all recording a lot of really excellent musicians.

Sgt. Pepper's was indeed made on a 4-track, but it sure wasn't made on a PortaStudio!

But it could have been. 

One thing to keep in mind, if you want to mix that way, and it works for you, more power to you. There really isn’t a right way to mix. There are so many different ways of doing it.

[triplets]

18 minutes ago, matthewbarnhart said:

True, but those records were also made by well-trained engineers who understood mic selection, mic placement, arrangement, analog gain-staging, tape recorder calibration, acoustics, etc... all recording a lot of really excellent musicians.

Exactly, they learned their craft, by just being there and trying out stuff and messing up.

Now we see this trend of "one click solutions" everywhere. Younger generations are afraid of messing up and learning over time. "Give me the solution now, I can't waste time on this".

[onerez]

1 hour ago, triplets said:

Exactly, they learned their craft, by just being there and trying out stuff and messing up.

Now we see this trend of "one click solutions" everywhere. Younger generations are afraid of messing up and learning over time. "Give me the solution now, I can't waste time on this".

Or…. Maybe some of us are at the end of our journey of life.  I never had the opportunity to be around the mixing and mastering…. I was just the guitar player.  I am not looking for a one click solution I am just trying to learn as much as I can….   Problem is this an art that could take years to master.  I am just maybe looking for some shortcuts. 😉. 

[wonshu]

I'm trying to have a "rough mix" ready where all faders sit at -6dB and I have a healthy output (with at least 9dB headroom) on my master. I use every tool available to achieve this: region gain, output in plugins or VIs or just a gain plugin in the channel.

The reason for me is, that we (in filmmusic) often have to work to the dialog and need to ride our volumes a lot. If that happens around -36dB or lower on the fader it becomes very cumbersome to automate and "draw" it as it jumps around quite a bit unless I blow up the track size to the whole screen. This is more a practical thing than it is an audible mixing thing but for our workflow getting healthy levels with the fader at -6dB is very practical.

[jmob]

9 hours ago, onerez said:

Or…. Maybe some of us are at the end of our journey of life.  I never had the opportunity to be around the mixing and mastering…. I was just the guitar player.  I am not looking for a one click solution I am just trying to learn as much as I can….   Problem is this an art that could take years to master.  I am just maybe looking for some shortcuts. 😉. 

Are you talking about yourself or, are you making a backhanded slight?

[David Nahmani]

4 minutes ago, jmob said:

Are you talking about yourself or, are you making a backhanded slight?

I'm pretty sure he's just talking about himself. 

[onerez]

3 minutes ago, David Nahmani said:

I'm pretty sure he's just talking about himself. 

I am talking about me….  I re read that and it does come off a bit backhand.  I am sorry.  I could have worded that better.   It’s mostly my frustration of trying to learn too much at once maybe.  I appreciate that I found this forum and they great info provided.   

[David Nahmani]

12 hours ago, JoshJ said:

is there a recommended Peak value to aim for into plugins then?

Not really.

Most Logic plug-ins work the same whether you feed them - 90 dBFS or + 90 dBFS. Obviously, I'm not suggesting you do that, for reasons other than distortion, which would not be created by using these levels.

The exceptions are obvious, plug-ins that have a threshold such as compressors or limiters or expanders etc... but then whether you use -18.2 dBFS or -18.6 dBFS also makes a difference on how those plug-ins will react.

And any plug-in that has a sweet spot (like an analog modeling plug-in or a guitar amp) needs to be toyed with to figure out the sound you get from one level or another. Engineers who use analog gear don't carefully gain stage to feed the analog gear exactly 0 VU, they often play with it and try to push it harder in order to get a specific sound that occurs only at certain levels, so if the digital counterpart have modeled that sweet spot then it's there to be experimented with until you get the sound you want. 

[David Nahmani]

9 hours ago, onerez said:

Problem is this an art that could take years to master.  I am just maybe looking for some shortcuts.

The art could take several lifetimes to master, so best to enjoy the process and just try your best. The technology we have today is easier than it's ever been though, so there's no need to overcomplicate everything, and that whole -18 dBFS calibration myth is definitely making everyone's life way more complicated than it needs to be. 

Basically you want to stay below 0 dBFS when recording and when bouncing. Anything else that happens inside the DAW is fine. But to make it easy to gain stage, what I typically recommend is that as a rule of thumb, your plug-ins are dialed in such a way that if you turn them off and on, that does not result in a huge volume jump. It should be about the same volume with or without the plug-in. Now don't take this as gospel and go back to all your mixes and painstakingly turn off and on every single plug-in while readjusting their level so that the volume is exactly the same: that isn't necessary, that's not a hard rule, just a rule of thumb, something to make your life easier. 

Let's say you're recording your vocals or your guitar. If your vocals peak somewhere between -12 and - 4 dBFS you're fine. You'd be fine if they peaked at - 1 dBFS (but that would probably mean that next time you record you risk going over 0 dBFS and clipping). You'd be find they peaked at - 20 dBFS (but that would make it hard to see their waveform or you may want to use the Gain parameter in the region inspector to compensate for their low level). 

Next, you insert an EQ, you should be able to turn the EQ off and on without having any drastic jump in volume. Next you insert a compressor, same, you obviously have the compressor affect the volume of the track, and probably make it louder when it's too quiet, or softer when it's too loud, but you want it to average roughly around the same volume, and not have to change the volume fader by 20 dB when you turn on the compressor. 

If you follow that simple guideline for most (or all) of your plug-ins then you've done great and there's no need to do any other gain staging. 

Hope that helps.

[JoshJ]

On 12/6/2022 at 12:20 AM, dubsak said:

Like Triplets says, Normalizing at -23 LUFS will work for a lot of things (think Bass, Guitar) but won't work well for others (Snares, Kick Drums or anything that's all about the transient). It's kind of why Normalize Region Gain doesn't work so well for music (as the source is all over the map unless you're dealing with mastered, complete songs or other material) so there's no definitive answer. You likely want to just watch your peak meters in the mixer and just see that peaks are in the yellow and the meter is mostly always in the green around -18 dBFS (again, it depends on the source). When people are saying things like "plugins are optimized or designed for 0 dB on a VU meter", that's what they mean. The best way to accomplish this (after the fact assuming you didn't gain stage into Logic to your satisfaction), is to use the region gain tool to raise or lower your gain per track by sight using those meters or the MultiMeter as Matthew recommended. Even with absolutely perfect gain staging, once you start using saturation, depending on the amount, you're going to increase the harmonics upwards quite a bit past 20kHz and there will be aliasing without oversampling. So…long story short, it all always depends. Kind of hard to have a definitive answer here.

 

This seems to have been what I was looking for in the end workflow wise. -23LUFS seems to get around -9 and -13dbfs peak (around 0VU) on most transient and tonal audio sources on the mixer meters. Thanks for the tip also re keeping an eye on the mixer meters so the peaks are in the yellow and the signal staying mostly always in the green around -18 dBFS. 

 

[JoshJ]

 

On 12/6/2022 at 12:20 AM, dubsak said:

You likely want to just watch your peak meters in the mixer and just see that peaks are in the yellow and the meter is mostly always in the green around -18 dBFS

Ok, testing this more on a new project. As you mentioned, when I normalize all my tracks to -23 Loudness, a bunch of the transient tracks are clipping but only peaking at -12 VU. So basically is the idea to use the Loudness normalization and then will still have to manually gain stage everything to around -18dbfs. Or cant i just normalize peak to -18dbfs in the first place for quick and dirty gain staging? Apologies if I’m still not ‘getting’ this..

 

Edited December 13, 2022 by JoshJ

[dubsak]

33 minutes ago, JoshJ said:

So basically is the idea to use the Loudness normalization and then will still have to manually gain stage everything to around -18dbfs

In this kind of instance, you've going to likely want to switch your normalization to Peak and set the value to -12. I think you're beginning to see that Normalize Region Gain offers very little help in terms of workflow if you're trying to just select all your regions and normalize everything at once. The most surefire way to gain stage your regions is to adjust the gain manually using the channel strip meters as your guide. The addition of the Gain Tool in 10.7.5 works a real treat for this: https://support.apple.com/en-ca/guide/logicpro/lgcp8e4c781d/10.7.5/mac/12.3#lgcpb57fdee6

Again, just adjust your gain so that peaks are around -12 in your channel strip meters. My suspicion is that you'll find this approach much more useful over the course of your music making. Hope that's clarifying.

[matthewbarnhart]

1 hour ago, JoshJ said:

Ok, testing this more on a new project. As you mentioned, when I normalize all my tracks to -23 Loudness, a bunch of the transient tracks are clipping but only peaking at -12 VU. So basically is the idea to use the Loudness normalization and then will still have to manually gain stage everything to around -18dbfs. Or cant i just normalize peak to -18dbfs in the first place for quick and dirty gain staging? Apologies if I’m still not ‘getting’ this..

Right, it seems like you still aren’t grasping the difference between Peak and RMS/average levels. Have you looked at Multimeter as I suggested?

A VU meter only shows RMS/average levels. Audio with lots of fast transients (drums, acoustic guitars, percussion, etc) have a relatively low average level compared to their peak value.

Look at a waveform of a snare drum and a distorted electric guitar and you can see the difference clearly.

In the analog tape world with only VU meters, we would usually try to hit 0 VU for bass, vocals, electric guitars, and synths; for drums, we’d start around -10 VU and listen for distortion, as the high peak levels weren’t represented by the VU meter. 

All that said, as many others have said, this is all an intellectual exercise in a modern DAW. It’s fine if you want to learn all this stuff (and there are many excellent resources out there — I started with the Yamaha Sound Reinforcement Handbook in the 90’s) but none of it will make a practical difference in your workflow today. If you just want to make music, forget about it. 

[David Nahmani]

6 minutes ago, matthewbarnhart said:

If you just want to make music, forget about it. 

This sums it all up nicely IMO. 

[dubsak]

3 minutes ago, matthewbarnhart said:

I started with the Yamaha Sound Reinforcement Handbook

This remains one of the best audio books out there.

[JoshJ]

2 hours ago, matthewbarnhart said:

seems like you still aren’t grasping the difference between Peak and RMS/average levels

 

Thanks. Yes, I understand peak/RMS average difference. My point is I guess, if the general aim is to be hitting -18dbfs ( @dubsak mentioned -12?) peak on the channel meters (with the aim of going into analog modelled plugs), why not just normalise to -18 peak? I understand it might not be optimal in terms of RMS but it seems like the quickest workflow if you need to gain stage 100+ tracks and you're not bothered about being exacting? Someone mentioned using -23 Loudness normalisation but as I found out, this doesn't seem to be an option as so much of the transient content clips using this setting. 

 

[triplets]

3 minutes ago, JoshJ said:

My point is I guess, if the general aim is to be hitting -18dbfs ( @dubsak mentioned -12?) peak on the channel meters (with the aim of going into analog modelled plugs), why not just normalise to -18 peak?

Because digital doesn't react like analog. You clip digital, and it will distort. You clip analog, and it will sound overdriven and warm.

Use -23 LUFS average loudness and start mixing!

[matthewbarnhart]

41 minutes ago, JoshJ said:

Thanks. Yes, I understand peak/RMS average difference. My point is I guess, if the general aim is to be hitting -18dbfs ( @dubsak mentioned -12?) peak on the channel meters (with the aim of going into analog modelled plugs), why not just normalise to -18 peak? I understand it might not be optimal in terms of RMS but it seems like the quickest workflow if you need to gain stage 100+ tracks and you're not bothered about being exacting? Someone mentioned using -23 Loudness normalisation but as I found out, this doesn't seem to be an option as so much of the transient content clips using this setting. 

Well there's your problem: the supposed analog modeling of these plugins is based on the RMS/average signal level, not Peak. Therefore, the general aim of all this gain staging stuff is to reach -18 dBFS RMS/average (0 VU), not -18 dBFS Peak.

To attempt to clarify further:

In your setup, 0 VU = -18 dBFS RMS/average. Not -18 dBFS Peak.

Analog systems typically used VU meters and RMS/average levels for measurements. (Analog peak meters exist, of course, but day-to-day studio work used VU meters.) But, transient-heavy material (drums) will have very large peak levels compared to the RMS/average values, so VU meters (aka RMS/average) were not as useful for these types of signals.

You can see this in the first attached screenshot from a Logic session of mine, where I've adjusted both the Snare drum and Guitar channels for an RMS level of -18 dBFS. (i.e. the assumed ideal for analog modeling plugins.)

As you can see, the Snare drum has a Peak level of +3.1 dBFS. In an analog system, this would clip the circuitry and likely cause distortion. (It will also clip the output of your converters, which you don't want.) The Guitar has a Peak level of -9.6 dBFS.

So, if you normalized all regions to -18 dBFS Peak, you'd get what you see in the 2nd screenshot: the guitar would have an RMS value of -26.4 dBFS, a good 8.5 dB lower than the "ideal" level for your analog emulation plugin. You'd be missing out on whatever harmonics and distortion they're trying to add.

Again, in a DAW none of this matters. It does matter in an analog system, because the closer you are to the ideal gain staging the better signal-to-noise ratio (and performance in general) you have.

And this is why I, a person who has 30 years of analog workflow ingrained in him, normalize drums to -6 Peak, and everything else to -18 RMS (or -23 LUFS, which is close enough for government work.) But a digital native really doesn't need to bother.

Edited December 13, 2022 by matthewbarnhart

[dubsak]

Part of this is how a VU meter works. The ballistics in a VU meter are pretty slow. So…for something like a snare drum, as @matthewbarnhart says, you'd look at something around -10 VU. That isn't to say that the peak of the snare drum hasn't gone higher. It absolutely has but a VU meter won't show that because it's very slow compared to a peak meter in your DAW (or a Dorrough meter). So, a tonal instrument like a bass or guitar makes using a VU meter much easier because its level is prolonged and the initial transients are not usually very extreme between say picking and holding the note. The ballistics of the VU meter, though slow, are a little more useful in this instance. If you set the normalization to peak at -18dB in Peak mode, a snare is not going to be the "-18" or "0 VU" they're talking about. They mean, as I think was mentioned earlier, that a sine wave at -18 dBFS = 0 VU. But that's a sine wave, not an instrument with a loud transient like a snare. But as several people have mentioned, this whole 0 VU think is not really something to be concerned about because of the nature of digital recording. In some instances, I believe it's purely marketing and in others, sure it's may be their reference but I think many people are fooling themselves into thinking that they hear or worse, the listener, hears a difference. Most people will say "sounds great!" not "I can really hear the subtle heat in the 1176 plugin you used. You really nailed the input."

And in the time I wrote that @matthewbarnhart followed up.

Also worth noting that I believe he worked at Electrical for a bit which is mostly to tape where all of this makes a difference. He'd be a valuable source about all this.

Edited December 13, 2022 by dubsak
Give cred.

[matthewbarnhart]

2 hours ago, dubsak said:

Also worth noting that I believe he worked at Electrical for a bit which is mostly to tape where all of this makes a difference. He'd be a valuable source about all this.

To be clear: I actually work at Chicago Mastering Service, which is owned by Bob Weston of Shellac, Albini's band mate. As such, most of my tape work these days is playing back 1/2" and 1/4" masters.

I do happen to live at Electrical Audio (long, weird story!) and work there as a freelancer from time-to-time. Both rooms are fully-analog (along with Pro Tools systems) and Steve still only works on tape.

The studio I co-owned for 20 years had both analog and digital capability, and most of the records I made there were 100% analog.

This is by no means an appeal to authority, but hopefully it's of use to the curious!

[JoshJ]

2 hours ago, matthewbarnhart said:

-18 dBFS RMS/average (0 VU), not -18 dBFS Peak

 

Ok. Believe it or not this simple fact hasn’t been clarified in any of the content I've come across on gain staging. People just say 0vu is = to -18dbfs 

2 hours ago, matthewbarnhart said:

But a digital native really doesn't need to bother.

The impetus of this post was regarding introducing unwanted aliasing from improper gain staging into modelled plugs. So still a little confused about that then if you're saying not to bother. 

2 hours ago, matthewbarnhart said:

normalize drums to -6 Peak, and everything else to -18 RMS (or -23 LUFS,

But this sounds like a good approach though. On that note, Hornet make a VU meter with a "Maximum" (peak protection) set to -6db by default . i think it adjusts the gain so that no peak exceeds the Maximum. Would this be useful in the case of drums then. For placing at the beginning of my drum bus?

 

Edited December 13, 2022 by JoshJ

[matthewbarnhart]

1 minute ago, JoshJ said:

The impetus of this post was regarding introducing unwanted aliasing from improper gain staging into modelled plugs. So still a little confused about that then if you're saying not to bother. 

Because often it doesn't make a difference.

Have you watched this video? 

 

 

Or this one? 

 

O

[JoshJ]

Thanks I'll take a look. I've been watching this guy a bit recently. He does a lot of A/B testing and alias visualising in Plugin Doctor and RX;

 

 

[polanoid]

30 minutes ago, JoshJ said:

Thanks I'll take a look. I've been watching this guy a bit recently. He does a lot of A/B testing and alias visualising in Plugin Doctor and RX;

That guy either doesn't know what aliasing even is, or he's deliberately spreading misinformation.