MEDIAART 2G03: Levels during Recording, Mixing, Delivery

In this page we're going to talk about levels in typical digital audio workflows. And we're going to structure our discussion around an idea about a typical workflow for a digital audio project that might be divided into three distinct phases. We might begin by recording or capturing some sounds, and then we might go through a long phase where we mix and transform these sounds, for example using digital audio workstation software. And when we're finished with all of that, there's a stage where we're going to optimize our results and put it in a form that is fit for delivery to other people and other listeners. And what we're going to see in the next few modules is that the management of levels works a little bit differently in each of these three stages.

Levels during Recording

We've already talked in other settings about the management of levels in recording, so this first module should be review a little bit. We know that we want to avoid clipping, because when our levels go above the maximum that we can represent, we basically lose that information and we get something that sounds very distorted as a result. We also want to keep levels well above the electrical and digital noise floors in our system. So we want the levels to not be above the maximum, but we also want them to not be sort of always close to the bottom either. And this leads to the idea of headroom. We do tests where we change our preamplifier gain so that our signals are coming in reasonably high (but not too high), where we're leaving enough headroom to compensate both for how unpredictable the thing we're recording is, and also how expensive or catastrophic it would be if we had to do it again.

So in the recording stage, to sum up, what we saw is that the ideal peaks of things that we are recording would be somewhere below the maximum (below 0 dBFS), but not too far below that maximum, which we accomplish by establishing headroom.

Levels during Mixing and Transformation

So imagine we move into a stage of a project where we're mixing and transforming (eg. editing) sound signals. And I want to show what typically happens here by moving to Reaper and describing what might happen.

So say I have an imaginary Reaper project, and I've got some tracks (I have five tracks), and let's say that their faders are in various different places because I've been mixing and I've been listening to the relationship between those things. And then I'm listening, and at some point the sum of all these tracks (all of those individual measurements getting added together) came to more than the maximum. Over here in the master area of the mixer, Reaper would leave an ominous red notation to warn me that the sum of all those tracks at some moment, came to more than the maximum. It clipped in other words, the sum of all these tracks clipped. There's nothing wrong with the individual recordings, we can see just looking at them that individually they're not clipped, but the sum of all of them together in our project clipped.

And so what we want to do in this situation is two things. We want to pull down the master fader a little bit. And when we do that, we're not going to change the relationship between our individual tracks in terms of level, they're going to have the same relationship to each other that they had before in terms of level, but the shape that results from all of them being added together, we're going to make that shape a little bit smaller. And say that I saw over in the master metre that I had clipped by about 3.4 decibels, then I would pull the main fader down 4 dB or more. And if I reset the metre and I listen again, hopefully this time it won't clip. In this way, I have some confidence that what I was hearing while I was working on my project is really the result of mixing all these tracks together, and not the sound of clipping and distortion from going over the limit.

To continue to describe that situation a bit more generally... when multiple tracks get added together, the levels tend to rise. And like in recording, we still want to avoid clipping. But unlike in recording, we can totally adjust and listen again when clipping does happen. So what we need to do when clipping happens during mixing or transformation, is we need to reduce the overall levels in the case of our work with Reaper by reducing the master fader, and we need to listen again. Because when we were listening before and it clipped, we were listening to the clipping rather than to the real mix of our tracks. And I want to emphasize that this is the only correct response in this situation.

What we don't want to do is we don't want to reduce those individual track faders, because those are for changing the balance between different tracks.

At the mixing/transformation stage, there isn't really an ideal digital level, but it is useful to be consistent or calibrated about our playback volume. So if we have speakers, it's very useful to kind of consistently have the volume or the gain on those speakers in a certain place. If we're listening on headphones, it's useful to have a consistent playback volume that's controlling what is sent to those headphones. It's useful to have that at a consistent place so we kind of know what to expect, but that's sort of a different question from our management of ideal digital levels. There isn't really an ideal digital level at this stage, we just need to do the right thing when clipping happens.

Levels during Delivery

And so then, let's say we do a bunch of mixing and transformation and we're happy with our project and it sounds good, yet our work is not done... There's usually a third phase of delivery or finalization, where now that we know really everything there is to know about our project, and now that we have total control over it, we just want to make sure that it gets to the end listener in as optimal away as possible. So this is what we're thinking about in the delivery or finalization stage of a project. So we have total control at this stage, and that means that we can make those final peak levels (whatever the highest point in those shapes that results from mixing all our signals together) whatever we need to make them.

So what do we need to make them? Well if we have a level, a peak level that's very very close to the maximum but not at the maximum, it'll be farthest away from that digital noise that we talked about in another module. And it'll also be farthest away from any electrical noise in the playback system where our file or where our result eventually gets listened to. And at the same time, by making this peak level not exactly at the top, we'll still be able to instantly tell this apart from a clipped recording. With a clipped recording, its highest level would be right at the top where it got cut off.

So that leads us to a useful standard for final delivery levels. It's not the only standard out there, but it's a very useful one and it's why we use it in this course. If we take our final digital audio results and we make sure that the peak sample is at -0.3 dBFS, what we'll be doing is keeping our levels far away from the noise in our system. And at the same time, we'll be producing something that is not clipped. We sometimes talk about this in other language, we might say that what we're doing is normalizing delivery levels to -0.3 dBFS.

Normalizing is that operation of taking a signal shape, and changing its size so that its peak or its highest level is at some particular place, and all the other levels are wherever they need to be to have the same shape. And so this leads to a final point in this module, which is about this operation of normalization more generally. So we've already seen that normalization to -0.3 dBFS is a very useful thing to do in the final delivery of a digital audio project, but we could do this at other stages too. However just because we could, doesn't mean that we should. In fact, normalizing individual source recordings in a project is usually useless, and sometimes harmful. The reason it's useless is because the individual sources in our project (ultimately the levels that come from those sources) are going to be set by our mixing decisions (we're going to make this one smaller and this one a little bit bigger). We do this I think driven by what we hear and by the creative possibilities that we hear in the material, it's not fundamentally a technical decision or something driven by numbers. And normalizing individual source recordings to a particular level can also sometimes be harmful if it's accomplished by making new audio files. Every time we make a new audio file, there's some amount of additional digital noise that is introduced. And in the same connection, normalizing anything at any point in the process to 0 dBFS (in other words, right to the top), makes no sense at all. But if you look at the files that you find out there in the world, you will have no trouble finding many files that have been normalized in this way. As the note on the slide says, unfortunately this is very widely practiced, but we're going to set a higher standard in this course. Normally we're going to normalize our material (our final delivered projects) to -0.3 dBFS, and basically we're not going to normalize anything else.