-18 gain staging using Supervision or other meters: max or average?

I’ve read through so many posts on gain-staging. I’m under the impression that initially setting tracks to -18 refers to Average RMS rather than true peak or peak RMS. Where in Cubase Supervision or other meters can I find a rough approximation to Average RMS signal level as opposed to Peak. I see moving lines in Supervision, but I find no explanation of those moving levels in Cubase help.

Or is my impression incorrect and people are using -18 Peak?

-18 dB isn’t some kind of special level that requires a high level of accuracy.

Back in the old analog tape days you had tape hiss as a noise floor and also a limited dynamic range. Because of this the practice was to mix hot - to maximize the use of the dynamic range and reduce hiss. It didn’t hurt that analog tape distortion added a desirable quality to the sound.

In digital recording all that is different. You don’t have any hiss or dynamic range to worry about so there is no advantage to mix things hot. And digital distortion sounds nasty so you want to avoid clipping at all costs so there is a good reason to mix at lower levels.

Folks aim for -18 largely because it keeps them a long way from potential clipping even if an inset adds substantial gain. You could successfully use -6dB as your gain staging level, but then you increase the risk of clipping. Likewise you could use -26dB to be extra ‘safe’ for no real benefit.

The point is -18dB is a bit arbitrary and you shouldn’t fixate on trying to hit it exactly. If the meter is bouncing around at about -18dB you’ll be fine.

The one caveat is that some plug-ins that model vintage equipment have input level sweet spots which they prefer.

It really isn’t critical, IMHO, exactly what levels you use these days on individual channels unless you are using plugins that emulate physical devices and expect an input of -18dBFS or so.

But it’s kind of traditional to use an average of around -24 to -18dBFS on each channel with peaks around -12dBFS or even occasionally -6dBFS.

I appreciate the replies. I know it is approximate. I suppose I can just keep peaks around -12, and I would be in the ballpark of -18 for average level.

I was also asking to understand the different moving lines in Supervision. Strange that I cannot find documentation of that.

Yes, I’m using a bunch of UA emulation plug-ins.

Barry

If you’re using UA’s plugins you can check their documentation for where they set their virtual “nominal operating level”. I think it’s -18dBFS average and perhaps one or two are different. Actually, I seem to recall this topic regarding UA coming up pretty recently. Perhaps check their forum.

As for supervision it’s going to depend on which module you are looking at. I believe both Nuendo and Cubase v12 has a VU meter. If you can set that so that 0VU=-18dBFS then all you have to do is look at the meter and if it’s bouncing around 0VU you’re averaging -18dBFS. Alternatively you can maybe use the loudness module’s “momentary” measurement. Short term is a little longer (I think 3 seconds) so “momentary” would be closer to VU.

What I’ve also done in the past is set the fallback time on meters in Nuendo so that they’re slower and resemble VU at least ‘on the way down’, though they’re still peak on the way up.

Just to reinforce good (IMO) replies here. Some use VU, some use peak level meters. I personally use peak level meters (directly the meters at faders side), aiming for -18 dBFS, but in an “eye-wise average” way, and allow for some rebel peak go up to -12 or -10 if they want. If other use VU meters or anything else, they will not set exactly the same levels (eye-wise average will never coincide with real rms measures), but they will have a good starting point, like I do, and that is what matters. You shouldn’t be aiming for rocket science-engineering exact values and you shouldn’t spend a lot of time at this stage. It is a rough starting point.

I have to say though that to me loudness makes far more sense than setting peak levels to some fairly arbitrary value. Typically we compress the crap out of audio these days so there’s really no way a very dynamic signal will survive the processing we do, a large range between average and peak is not really a problem. If average is at -18dBFS and there’s a peak at -3dBFS for example then I’d say that instrument or source, whatever it is, is usually not likely to be that dynamic after we’ve slapped a bunch of dynamics processing on it.

And to add to that we’re in floating point processing so unless there’s a problem with a plugin we won’t clip until we hit the converter, which leads to room calibration and reference levels and 0dBFS being very loud…

So to me from the standpoint of balancing an average gives a better starting point and also corresponds better to analog emulations since they’ve aimed for a nominal operating level which is more along the lines of average than peak.

But it’s not like there’s a rule on this so whatever works I guess…

-18 is average so that allows for peaks to go much higher but not clip.

Agree with the posts here, but would also like to add that for very sharp percussive sounds, aiming at -18db avg can be quite hot and sometimes close to 0dBFS. For those sounds looking at a peak meter is more useful.
My own rule of thumb is roughly -18dB average/0VU or -10dB FS, whatever comes first.

2 Likes