Gain staging question


I’ve understood it’s important to do proper gainstaging to avoid digital clipping and to hit the sweetspot of analog emulation plugins. Let’s say I put a general input level of -18 dBFS on my tracks, should I do that on my buses and mixbus as well?

For example if I want to use the J37 Tape on mixbus, I guess the VU level should be around -18 dBFS to hit the sweetspot. But then it seems like quite a lot of headroom before mastering. Perhaps not a bad thing but I’m curious how others are doing this cause I usually read about pre-mastering headroom around 3-6 dBFS.

In Cubase and Nuendo you’re not going to clip any of the internal routing/summing, ever. The only point you might clip your signal is on your master output because it usually needs to convert from floating point processing (more headroom than you need) to fixed point (i.e. 16-/24-bit). So the output should be below clipping.

If you have an audio track that’s above zero going into a group that then also is above zero going into an output that’s above zero then at some point you need to lower the level. If you do it at the output it should be fine even though the earlier stages are above zero. It’s generally not a productive way of working so it’s best to maintain levels throughout the chain so you don’t have to do this, but it should be fine.

As for plugins:

The word “sweetspot” is a bit misleading I think. Analog hardware has a nominal operating level where everything is working as linearly as possible (or “as intended”) and then there’s headroom above that, and then at some point you start getting some sort of non-linear behavior which we can call “distortion”.

So if you were working in a professional studio decades ago and recorded to tape you’d hit that tape at 0VU which is nominal operating level and the signal would be clean. If you hit the tape harder than that you started to get distortion of the signal. Some people like that, and in one sense even at the nominal ‘clean’ range of the hardware there was some amount of change to the signal - hence the ‘sound of tape’.

Now, if you’re using a plugin you should ask yourself why you want to use it? If it’s because you want a subtle tape sound then you should hit that nominal operating level. You should get some minor amount of tape sound I would imagine. But if you want “more” tape sound then you have to hit the plugin harder to get more distortion. It’s all aesthetic and not technical. Basically you can simply start at -18dBFS average and listen to it. Adjust until it sounds the way you want. That’s it.

As for the output level perhaps one way of thinking about it is that if you want it to be louder you hit the tape harder. The non-linear distortion will likely limit the dynamic range of your signal, i.e. ‘compress’ it, and that’s really what you want if you want to make it louder. So it actually makes sense to hit that plugin harder if that’s what you want and you can then ‘forget’ about -18 (unless you have another plugin after this that also has a nominal operating level in which case the output of the first - or the input of the second - has to be adjusted).

And finally you should of course read the manual of the plugin. What does it say about nominal operating level and distortion?

PS: “Sweetspot” is a bad word in my opinion because to me it has to do with the intent of someone. In the past the “sweetspot” of the engineers that designed and built the tape machines was where it operated cleanly with low distortion and noise. Today however when we want that sound the “sweetspot” for you might be far higher, because you like a certain amount of distortion. So the word to me is more subjective these days and therefore less useful.

1 Like

@MattiasNYC Thanks for your extensive reply. Does this mean that I don’t have to care about if my eq or compressor is clipping internally as long as it sounds good and the master output is not clipping?

If we take another example which don’t have an input control, like Waves CLA-2A, will the gain staging affect how it sounds? According to the manual the Peak reduction is non-linear, but I’m not sure if that only affects the amount of compression or also sound coloration.

Well, if it doesn’t clip then there isn’t a problem. I don’t actually know which plugins can and do clip or if it’s audible. If you’re worried about this the easiest thing to do is to set up a very simple signal chain and test it. Just create an audio track, import some audio onto it, pick a plugin and then see if you can make it clip to the point of distorting - as you’re doing this you’ll want to turn the output down to compensate so that you know that anything you hear that is a problem is not coming from the master output clipping.

I honestly don’t know enough to say if plugins can clip or not, but my intuition says ‘yes’. Whether or not you’ll hear it or if it bothers you is a different thing. But either way, just try it.

Well, if I was Waves then I wouldn’t call the peak reduction non-linear if it only meant that it compresses more the higher the input is. I mean, I would sort of expect that to be the standard way any compressor functions. What I would expect instead is that as that signal compresses it would get more distorted the more compression is applied, or the higher the input level is.

So ‘yes’, I would expect a properly modeled LA2A compressor to not compress linearly, and instead change the character of the sound and not just the level as the input level or compression level increases.

This too you can test by essentially doing the same thing. Just play back audio through it and try different input levels and compression levels and just make sure to compensate at the channel’s fader/output.

– I would also say that plugins like that may never technically “clip” in a digital sense and instead just do it as a result of the desired (analog) model. So in other words you may reach a ceiling of sorts of that plugin, but it isn’t a limitation of digital but a designed effect. I suppose we could call that “clipping” but that would maybe be confusing because it wouldn’t be the same issue as a bunch of 0dBFS samples in a row.

1 Like