Proper Gain Structure & dbfs?

[MyKill]

Hi all. Sorry if this has been answered previously, but a search of our fine forum using combinations of these key words yielded no results. If there is a post answering this, please just link me to it (and keep you snarky remarks to yourself )

If it matters, Cubase 6.0.4, custom DAW computer, 8GB RAM, AMD 6 core 3.2 gHz processor, and its punk/ rock. Also, unless otherwise noted, all references to levels will be in dbfs.

So, I'm not a noob, but perhaps this is a noob question. What is the proper gain/channel structure for mixing ITB?

To be specific, I mean,is it better, mechanically inside the DAW to (just an example, not stating this is what I do or these numbers are goals) have my snare TRACK peaking at -6 and pull the DRUM BUS fader way down so the snare peaks at around -18 to account for all the summing taking place,

OR

have my snare TRACK fader down so it's peaking at -18 and leave plenty of head room to start with in my DRUM BUS so the bus fader pretty much sits at unity when not being automated?

I know that all the Cubase faders have better resolution for -12 to 0 than from -∞ to -12, but in terms of available headroom and signal to noise ratio, is there a "better/preferred" method for fader levels in the mixer for the signal path?
Is one way better than the other, or does it even matter?


Let me explain as maybe I'm not using the correct terminology:

I want to keep my pre-master, mix level (in dbfs on the stereo output meter) at around -3 to -6 dbfs (according to some).
So obviously, my individual tracks and busses cannot be hitting anywhere near that as the total sum of all tracks must peak at ~-6.


I have Googled this question for the better part of a week and found people strongly preaching opposing views in terms of "starting level". Some say keep your snare at -6 and build around that, so people say "use your ears" and "whatever sounds good"

The whole "use your ears" & "there is no "right way" thing frustrates the $#!T out of me. I went to school for this (although obviously didn't pay enough attention to this point), but the reality is, there is a rough set of "better/best practices" used by successful pros. I understand that your ears are the best and most powerful tools you own; better than any spectrum analyzer, VU meter, dbfs meter, book you can read or tutorial you can watch. I GET IT, OK?!?

But, "usung your ears" as a catch all answer is absurd. Some people have better ears than others, whether through good DNA or through years of training. "Use your ears" is good advice, but usually after you understand the Fletcher-Munson curve, have a firm working knowledge of the frequency spectrum, a reasonable grasp of psychoacoustic principals and lots of hours logged mixing different genres, finding what EQ's, compressors, etc. work the best on what instruments/styles. Maybe to MY ears, slamming every buss channel with -25 db of compression sounds good (it doesn't, but it's just an example), but that doesn't mean the general public who is accustomed to radio production will think it sounds good, and that most engineers/producers wouldn't rip me apart for killing every transient in the mix!!!!

So if you are going to respond with the one sentence "use your ears", then GFYS.
Ok, disregard what I said about snarky remarks earlier

[jupiter8]

So, I'm not a noob, but perhaps this is a noob question. What is the proper gain/channel structure for mixing ITB?

It does not matter.

[jupiter8]

Double the fun.

[MyKill]

It does not matter.

It's got to matter a little as far as bit depth and dynamic range though, right? I do not claim to have wrapped my head around this subject completely yet, but in that last -6 to 0 dbfs range, you are adding those bits and doubling the dynamic
range, correct?

So if I take my signal (in this case a snare compressed on the way in and with a pretty consistent level of ~ -1.2) , leave it's fader alone, output it to the drum buss whose level I adjust so the snare sits at , say -12, I then have all that dynamic range to send to effects and buss compression and what have you and keep the Drum bus output eventually as ~ -9.

You're saying there is no difference in the final mix if I keep my actual snare fader low, say -18, go through the same bus and fx chain, which would mean eventually adding gain to the buss signal to get it to -9?

If so, is that because the recorded signal itself has peaks at -1.2, therefore making full use of available bits and thus dynamic range?

Am I looking at this the right way?

[Aroused by JarJar]

I don't know the answer to your question. I just do like in analog- power should be upstream, flows downstream (as much as possible).

[MyKill]

I don't know the answer to your question. I just do like in analog- power should be upstream, flows downstream (as much as possible).

As someone who has only seen a reel to reel once, has no VU meters or console and learned everything about production ITB, can you elaborate on your river metaphor?

[jupiter8]

It does not matter.

It's got to matter a little as far as bit depth and dynamic range though, right? I do not claim to have wrapped my head around this subject completely yet, but in that last -6 to 0 dbfs range, you are adding those bits and doubling the dynamic
range, correct?

So if I take my signal (in this case a snare compressed on the way in and with a pretty consistent level of ~ -1.2) , leave it's fader alone, output it to the drum buss whose level I adjust so the snare sits at , say -12, I then have all that dynamic range to send to effects and buss compression and what have you and keep the Drum bus output eventually as ~ -9.

You're saying there is no difference in the final mix if I keep my actual snare fader low, say -18, go through the same bus and fx chain, which would mean eventually adding gain to the buss signal to get it to -9?

If so, is that because the recorded signal itself has peaks at -1.2, therefore making full use of available bits and thus dynamic range?

Am I looking at this the right way?

You'd be correct if the DAW is using Integer maths but all modern DAWs nowadays use floating point. The main reason for that is that gain staging is a non issue.

Now there are good reason for "proper" gain staging but those are workflow related, it's got nothing to do with sound quality.

Just for funsies i once lowered the volume of a sample with 120 dB,rendered it as 32 bit float. Increased the gain by 120 dB and it canceled out with the original completely IE they were exactly the same.

[Aroused by JarJar]

I don't know the answer to your question. I just do like in analog- power should be upstream, flows downstream (as much as possible).

As someone who has only seen a reel to reel once, has no VU meters or console and learned everything about production ITB, can you elaborate on your river metaphor?

In analog the rule of thumb is to turn things down as the signal flows to its final destination, not turn things up. At least that's what I learned decades ago, and in my experience it's correct.

That's a general rule and not a law, however. You might want a radical EQ boost of the highs going into tape or something like that, and there are musical concerns that might violate that. For example, recording tiny little quiet sounds and boosting the hell out of them can make huge sounds, and I'm sure when it comes to electric guitar amplification and effects there can be gainstaging within the chain that violates the rule.

[cron]

Bear in mind that any analog modelled plug-ins you own will respond differently when fed different input levels and so may benefit from analog style gain staging. The same is true for some pure digital effects where things can be far less subtle, bitcrush being the obvious example. Also, some plug-ins (Satin being a recent example from what I can gather) may have a 'noise floor' of their own that you might want to avoid boosting later on. Other than that, the only reason to gain-stage in a DAW is for workflow reasons. You can whack up the master at the end to compensate or render to 32 bit float then normalise/boost. I prefer keeping things low so I never have to worry about headroom during mixing.

[DuX]

I think people are making nuclear science out of gain staging and the more complex the explanation - the more confusion.

It is actually so simple: just keep the volume of all the tracks in the mix around, but not beyond, -18dB average RMS, and use the VU meter for that. It's that simple. Also, keep the same volume between the plugins. That's it. I use Satson CM for that. I love that plugin so much. Except the stupid reflections on its "screen"... but that's not relevant for this thread...

I keep my tracks at around -18dB average RMS and usually finish up peaking at about -14dB average RMS on the master. Usually I get to around -3dB dBFS peak on the master channel this way.

[Compyfox]

The whole "use your ears" & "there is no "right way" thing frustrates the $#!T out of me. I went to school for this (although obviously didn't pay enough attention to this point), but the reality is, there is a rough set of "better/best practices" used by successful pros. I understand that your ears are the best and most powerful tools you own; better than any spectrum analyzer, VU meter, dbfs meter, book you can read or tutorial you can watch. I GET IT, OK?!?

The thing is, the "old rules" (which were great rules btw) were thrown out of the window in the mid 90ies, while digital machines took over, and more and more rules were bent to gain as much loudness as possible.

Nowadays, it really doesn't matter anymore what you do, with a big huge emphasis on the word "but".




In the last couple of years, more and more software emulations came up. Nowadays, pretty much all of them use a reference level(!), and that reference level is usually from the hardware days - ranging from -24dBFS, to the more common -20/-18dBFS, to sometimes even -15dBFS or -12dBFS for tape machines - resulting in a much lower headroom for peaks in return.


So what's the "optimal" way?
Use both a digital peak meter and a VU meter.
Use color codes for your digital meter


Simple rules, huge effect if you want to use emulations properly, integrate hardware, have a better fader precision and not touching the master fader at all (or even use a comp/limiter on the summing bus).

The color codes I recommend for channels are:
Green - up until -18dBFS
Yellow - from -18dBFS to -9dBFS
Red - from -9dBFS

How to use these codes:
Level in your signal so that transient intensive material (i.e. snare) doesn't exceel -9dBFS. If you have bass intensive material (like kick and bass), use a VU calibrated to -18dBFS (the crossover point from Green to Yellow, and apparently an international standard in terms of reference levels). Then level in the signal to either hover around -18dB RMS or 0VU (which is the same btw).

First gain stage done.


Now you can easily mix to your hearts content, but don't try to exceed -3dBFS on the summing bus.




If your ADC/DAC uses +4dBu (Studio Level) as input and output, you can even simply integrate outboard gear that uses the same worklevel and is also calibrated to -18dBFS = 0VU.



There you have it, use whatever bitrate you like, use whatever software/hardware you like. Problem free zone.

Hint:
- my KVRmarks
- this YouTube video: http://youtu.be/TGFy-s-mG70

[fese]

Well, Compyfox explained the theory better than I (slightly advanced noob:) could, so I'll just chime in with how I do that in practice.
I use VUMT set to -18db in all channel to set up the gain according to the rules described by Compyfox, with the Cubase channel faders set to default. VUMT is great for that, but there are other utilities one could use.

First gain stage done

Then I use the faders to balance the tracks.
I cannot really prove this or argument theoretically but I think my mixes tend to sound better since I work like that.

Never heard about a digital peak meter where on could set colors, though. Compyfox, do you have examples?

[Compyfox]

Cubase 6 and 7 can do that in the program preferences. I think zplane's PPMulator is also capable of that, so is the Nugen Audio Vis-LM, and pretty much every digital meter with editing capabilities.


The color codes for a digital meter are heavily inspired by PPM meter specs/limits like the IEC 60268-10's and IEC 60268-18.

[Jafo]

Some compressors and ampsims like to see a certain dbfs due to inflexible input gains, but it's easy enough to fudge things by adjusting levels before and after with volume filters by ear.

[NAD]

Just for funsies i once lowered the volume of a sample with 120 dB,rendered it as 32 bit float. Increased the gain by 120 dB and it canceled out with the original completely IE they were exactly the same.

I bet you're a lot of fun at parties!