Gain staging for plug-ins. Freedom from gain staging. Gain staging for your final mix and master. Hi, I'm David Mellor, course director of Audio Masterclass and in this Sound Sound podcast, I'll be covering all of these topics. To get the best from this podcast, which is episode four in my gain staging series, you will need to listen to episodes one, two and three but don't worry, with modern digital audio workstation software there's hardly anything to go wrong, but do go back and take a listen when you have chance. Gain staging is an interesting and important topic and it's good to have all of your knowledge and skills in place. So you've listened to my earlier episodes and I don't have to re-explain every tiny detail. OK, let's get started with gain staging for plug-ins.

Gain Staging Your Plug-ins

As I said previously, you'd have to be doing things in really rather crazy ways to upset your DAW. Just don't clip when you record and don't clip when you mix. But what comes in between can be interesting and by in between, I mean plug-ins. Plug-ins come in two types, those that are purely digital and do the job to the best that modern digital technology can achieve and those that, although they are digital, they emulate traditional analogue hardware and this does make a difference for gain staging. Your digital audio workstation software has amazing internal headroom. In fact, you'd find it difficult to clip your DAW internally. However much you boost the level, you can always bring it down again and it will be clean - well, up to a point. But believe me, you'll never reach that point unless you seriously set off on an expedition to find it. So plug-ins that are purely digital can have amazing headroom too. Maybe not quite so much as your DAW itself, but very good nonetheless.

I can set up a demonstration here. I'm going to use Pro Tools, but you should try this test in whatever digital audio workstation software you use. Try it with a plug-in that doesn't seem to be emulating analogue equipment, but is purely digital in intent. I'm going to load up the basic Pro Tools 7 band EQ and explore the limits of its headroom. Firstly, my favourite frequency of 220Hz, the A below middle C, because it'll work just fine and it isn't too hard on the ears. By the way, I'll have to lower the levels for inclusion in this podcast, but all of my demonstrations will be in proportion and none, hopefully, will be too loud or too unpleasant. Here's the tone with no plug-in, 0 dBFS. And now I'll put it through the EQ with neutral settings. OK, that's what we expect. Now I'll boost the level by 20dB before the EQ, so that's +20 dBFS, which the DAW can handle easily, but can the plug-in handle it? Then I'll lower the level after the EQ by 20dB to bring the level down to 0 dBFS. All these demonstrations are for real, by the way. No smoke, no mirrors and you can try them for yourself.

Here's the tone, +20, EQ, -20. It's clean. So this plug-in clearly can handle levels up to +20 dBFS. But can it go higher? Here's +20.1 dBFS, again with - 20.1dB after the plug-in. Oh dear, that's harsh. So we now know the limit of this plug-in. Seriously, I don't see why it can't be higher. I suspect that if I tested other purely digital plug-ins, not analogue emulations, I'd find higher headrooms in some. But 20dB above 0 dBFS is good, but there's something else to consider. This is an EQ and I've been using it flat. What happens if I apply some boost will that eat into the headroom? So I'll take my 220Hz tone at 0 dBFS and send it to the plug-in. I'll use two bands, both centred on 220Hz and apply a total of 20dB of gain at that frequency. I have to use two bands because the highest one band will go is +18. Again, I'll lower the level by 20dB after the plug-in, so we will be back at zero. Will it be clean? Yes, it's clean. I'm going to spare you the distress and just tell you that if I go even a tenth of a dB higher, I get harsh clipping as before. What we've learned therefore is that this plug-in can handle inputs up to 20 dBFS, but any boost inside the plug-in will cut into this headroom. I think we now know all we need to about gain staging this plug-in. The headroom is good, very good, but not infinite. I tried this test with a couple of other EQ plug-ins, the SSL Native X EQ 2 and the Slate Digital Infinity EQ. Same results. Testing is good.

OK, let's move on to an analogue emulation. More or less at random, I've picked the Rule Tec EQ1A, which is an emulation of the famous Pultec EQP 1A, which, as you know, contains valves, vacuum tubes, thermionic tubes, warm to the touch and warm to the ears. So with the real life Pultec we would expect some warmth even at lower levels and we would expect the degree of warmth to increase as the level is increased, to the point where the sound will become harsh. So, what happens in the Rule Tec? I'll give you a range of input levels and in my demonstrations I'll always correct this after the output, so the levels you will hear will be the same, but the sound texture may, indeed should, vary. I'll make these tests with the settings in their neutral positions, EQ flat in other words. Oh, a complication. This plug-in boosts the level by 2dB, even with neutral settings. I'll compensate for this and we'll just pretend it doesn't.

So let's start low at that famous level of -18 dBFS, which in the physical world would be 0 VU. Yes, there is some warmth in this. I can't hear much, but I can definitely see it using a Spectrograph plug-in. Let's go up to -6 dBFS. A little bit more warmth perhaps. In the real world, this would be +12 VU. So clearly, if this plug-in is an accurate emulation, the real Pultec has some useful headroom. Let's cut to the chase and find the limit, +20 dBFS, which is +38 VU. I think in the real world, the meters would catch fire by now. This is getting very warm, hotter than July perhaps, but still this amount of warmth could be useful, particularly for drums, but +20.1. OK, we've found the limit, curiously +20 dBFS again. I'm sensing a pattern here.

Of course, I can't test every plug-in in the world and neither can you, but we can test our own plug-ins to know where the danger zone lies. What we can see from these two tests is that the purely digital plug-in is clean all the way to +20 dBFS. The analogue emulation displays increasing warmth as the input level rises, then out and out clipping at +20.1. All of this applies to other types of plug-in too. Compressors will be interesting because their whole purpose is to control level, so I suggest you experiment with your favourite compressors. Find their sweet spots and find their limits. If you see a red light, trust your ears instead.

32-Bit Float

OK, moving on. As I've said several times during this podcast series, because it's important, you must not clip while recording and you must not clip when mixing. But inside your digital audio workstation, you have loads, I mean loads of headroom above 0 dBFS. In a 16 or 24 bit WAV file, you have none.

But what if there were a file format that could handle the huge range of levels that are possible inside the DAW? And there is such a thing, it’s called 32-bit float and it can be stored in a WAV container. So, if someone sends you a 32 bit WAV file and you import it into your DAW, don't be surprised to see red lights all over the place, because it may well have levels way above 0 dBFS. Which, of course once you realise, you can compensate for and everything will be fine. So, how does it do this? Well, keeping this simple, a 16 bit regular file chops up the audio into 65,536 different levels and stores them as binary numbers, 44,100 or 48,000 times every second. 24 bit files have over 16 million different levels. So, you can expect a 32 bit file to have 4,294,967,296 different levels. Phew.

But it doesn't work like that, it’s much more clever. This is where the float part comes in, short for floating point. Keeping this simple, a 32-bit floating point file uses 23 bits to store level data, 1 bit to show whether the value of any one sample is positive or negative and the other 8 bits to scale the dynamic range up to 1528 dB. This is either madness in a sane world, or sanity in a world of madness. But whoever thought of this idea for audio has certainly given us what we need and more and more and more. Maths is fun, isn't it? But what about the practicalities - what can 32-bit float do for us?

It starts with recording. There are digital audio recorders that record in 32-bit float. An example of this would be the Zoom F6 field recorder, which features dual analogue to digital converters translating to 32 bit floating point recordings and the dynamic range is so vast, it doesn't need a gain control. Think about that for a moment. In audio, we've been setting the gain ever since time began - audio time. There has never been a recording made without someone setting the gain control. OK, there is such a thing as automatic gain control, but they never really worked properly, did they? And they were anyway just doing what a human otherwise would, just not as well.

I could analogise this with photography. When is the last time you focused your camera? I'm old enough to remember when autofocus was first invented and I bought an autofocus camera as soon as I could afford one. Freedom! Freedom to compose the shot and not worry about an irritating technicality. And now we have the same freedom in sound. You'll never have to worry about clipping your recording ever again. And headroom, well, there's so much headroom, you never have to think about it. 32-bit floating point recording devices are still thin on the ground at the moment, as are 32-bit float audio interfaces but just wait, give it a couple of years and there won't be anything else.

Mixing And Mastering

So this takes care of recording. Your recording is perfect without any possibility of clipping. Your DAW has a similar dynamic range, so no problems there. You could still, as we have seen, clip your plug-ins, so be careful there. What about the output? Your mix and master. So, we have some other things to think about here. Yes, you can mix to 32-bit floating point. I can do it in my Pro Tools and you can probably do it in whatever modern digital audio workstation software you're using. But as the saying goes, just because you can doesn't mean you should. 

OK, let's back up a bit. During the production process it's common to want or need to bounce instruments or vocals individually or in combination. An example might be that you've recorded a backing track and you want to send a stereo mix of that to a colleague who will add the vocal, bounce that and send it back to you to incorporate in the multitrack. This is a perfect use case for 32-bit float. It's a why not. If you bounce to a 24 bit file, you have to worry about clipping. Well, it's not that hard, but it is one more thing to think about. With 32 bit float, you just bounce, that's it. My view, therefore, is that anything that involves bouncing during the recording process should be 32-bit float, particularly if it involves collaborators. There's just no reason not to use it. I should say that if your audio interface is 24 bit, and you play a 32-bit float file peaking higher than 0 dBFS, you will hear distortion. It's in the interface rather than baked into your project, so there's nothing to worry about, just use your fader and bring down the level.

But what about the mix? No, let's skip to the master. What about the master? Should you master in 32-bit float? No, no, no, no, no, no. I said no. Did I make myself clear? No, don't do that. Let's suppose that you're going to sell digital downloads on your own website. You're perfectly able to make 32-bit float files and sell them. Your customers will probably be able to play them. I've just checked that by playing a 32-bit float file on my iPhone. It plays just fine. So what's the problem? It's the loudness war. I'm pre-empting myself a little because I'm going to talk about loudness when I come on to gain staging for mixing and mastering. But you know what the loudness war is. It's how in the 1990s label bosses and mastering engineers realised that although 0 dBFS may be the limit, it’s possible to increase subjective loudness by compression, brick wall limiting, and harmonic enhancement. That's enhancement in air quotes. So for the last almost 30 years mastering engineers have been falling over themselves to make louder and louder masters and label bosses have cheered them on. We want more. I'll say more about this later, but for the moment we can consider that the loudness war is a thing and despite the best efforts of streaming services, it hasn't gone away yet. Label bosses want more loudness and mastering engineers, if they want to stay in their jobs, have to comply. I mean no disrespect to mastering engineers by the way, it’s just the way things are and loudness pays. So it seems.

OK, things are bad. Can they get worse? Yes, they can. They can get a whole lot worse. With conventional 16 bit or 24 bit files, fixed point we can call them, though I haven't used that term until now, there is the ultimate horizon of 0 dBFS. There's no going any higher than that. With 32-bit floating point, there's virtually no limit. If 32-bit float mastering came to be a thing then at first it would surely be a benefit. But then the quest for more and more loudness would start. How bad can it get? Where will it end? I said I'd played a 32-bit float file on my iPhone. It plays fine. Well, I also took the opportunity to boost the level, so that from my 24 bit audio interface, the clipping was extreme. On my iPhone I could turn down the volume and it was fine. Mark my words, if 32-bit floating point mastering becomes a thing, people are going to exploit this and it will not be for the better. 

That was a bit of a rant, wasn't it? 24 bits are fine for final masters, absolutely fine. We don't need anything more. Danger lurks that way. One more thing. I said, suppose you're selling your tracks so you could, if you wanted, sell them in 32-bit float format. Well, you probably want wider distribution than that, so you should submit your masters in whatever format your distributor wants. For Apple digital masters, the recommendation from their own documents is a 24 bit file. For CD Baby, for example, 16 bits. But if a distributor specifically asks for 32-bit float, I'd say don't peak over 0 dBFS.

Gain Staging The Mix

Mixing and mastering - well I’ve covered this in part already, but it's a whole topic in its own right, so I need to treat the end game of the recording process with full respect. So how do you gain stage your mix at the point of output? I could say simple, just don't clip. But as with a lot of things in life, there's more to it than that.

Firstly, there's the matter of inserts. I'm going to speak from a Pro Tools point of view. Other digital audio workstation softwares are available and they are capable of equally good work. You may have to conduct your own investigation into their behaviour, but my points will apply in concept to all DAWs, in fact all audio. As you know, your audio and instrument tracks have inserts where you'll put EQs, compressors, harmonic enhancement, delay, transient shaping, then a couple more EQs and compressors, well, that's what they do on YouTube, so it must be good. No! Do whatever you like that sounds good, it will sound good before the fader, then you'll control the level of your track with the fader and it will sound the same at any level, just louder or quieter. All of your tracks will be like this. It's so sensible that we hardly ever talk about it because anything else would be lunacy.

But the master track, master channel, or master fader, whatever you like to call it, is different. The inserts come after the fader, so whatever you do to the fader affects the plug-ins you have inserted into the master. Does it affect all of the plug-ins? Not really. A plug-in that's totally clean should be unaffected, unless you've gotten everything I've already talked about gain staging wrong. But a compressor? Yes, it is affected hugely. A limiter, yes. Harmonic enhancer, yes. Analogue emulation, yes. They're all affected. Changing the position of the master fader doesn't just change the level, it changes the sound. There is logic to this, I don't mean Apple's logic, just everyday logic.

If, for instance, the inserts were pre-fade and you inserted a brick wall limiter or a dither plug-in and then changed the setting of the master fader from 0dB, you've just ruined your mix. Your brick wall level will be wrong and dither will either be too loud or not do its job of removing digital distortion properly. But for everything else, you know, I personally would be quite happy if everything was pre-fader. I can do this however by mixing everything to an Aux track and placing my inserts there, then bus that to the master and have just my limiter in that track and dither if I need to master to 16 bits. Or I can put a trim plug-in after all my sweetening plug-ins and before the limiter and dither and set my final level with that, leaving the master fader at zero.

Of course by the time you hear this, Pro Tools might have updated to allow a pre-post option. If there's one already I can't find it in the settings, the manual, or anywhere on the web that Google can find. So, that's it for that. Once you know that your master fader has post fade inserts, you'll know how to handle it and for DAWs other than Pro Tools, I strongly recommend you check.

Headroom

Now, how much headroom should you leave in your mix? This is a contentious question. I'm happy for people to have different opinions from mine, but I'm going to give you mine. I am, of course, right. Or am I right? That will be for you to decide, but please hear my arguments out. In no particular order, one school of thought is that digital signals degrade when they're close to 0 dBFS. In my humble opinion, they don't. In fact, they get better. Better signal to noise ratio, lower distortion. The sound of analogue degrades at higher levels, but that's the olden days. We've moved on.

Next, a commonly heard phrase is, leave headroom for mastering. OK, suppose you've mixed all the way up to 0 dBFS. You haven't clipped and if you looked closely at your waveform, you would see that there are never two or more consecutive samples at 0 dBFS. It's clean. You send your mix to your mastering engineer, but it's already maxed out. What can they do? I really don't know where this issue has come from. Of course the mastering engineer can turn up or turn down your mix in whatever way they find convenient. There's absolutely no problem here unless you've clipped or your level is ridiculously low. Where I could imagine this comes from is that one of the expectations of mastering is that it will make your mix louder. Yes I know, loudness war, bad thing. But it is a common expectation. So if your mastering engineer asks you to leave 6dB of headroom at the top of your meters, then your master will come back louder and as everyone knows, louder is better. Where's that sarcasm emoji when I need it?

There is the genuine point, however, that the mastering engineer really, really wants you to avoid clipping. I've received many pieces of audio over the years that were clipped, some from people who should have known better. When a mastering engineer works day in, day out on other people's mixes, it's great to have a guarantee that none will be sent in clipped. So in that sense, a few dB of headroom isn't a bad idea. Pro tip, if your mastering engineer suggests you leave headroom, just give them what they want. It'll make life easier and when you get the master back, compare it with your mix at the same level. That way you'll know whether or not the mastering engineer has improved your sound.

Mastering Using LUFS

We're on the last lap now. You're mastering your own work. Never a good idea but most of us do it anyway. Yes, mastering is all about getting the stereo mix to sound at its best, but it's also about setting the correct output level. This will be your final gain staging decision before your track hits the streets and so we have loudness units relative to full scale. L U F S. LUFS. This harks back to TV commercials being louder than the programmes. Advertisers think that if their commercials are really loud, you'll buy their products. Well, I can't think of any other explanation. This isn't so much of a problem now because broadcasters are mandated to control their loudness levels of programmes and commercials.

So in a nutshell, what are LUFS and how are they measured? Well, there's a whole Wikipedia to answer this in detail, but I can summarise that they respond to loudness pretty much like a VU meter, which is to say, pretty much like the human ear. It's how loud the audio sounds rather than what its sample peak or voltage peak is and measurements come out in LUFS, which technically I suppose is a ratio, but practically we can consider it a unit.  LUFS measurements are nearly always negative. I thought they were always negative until one of the denizens of YouTube made a track with positive LUFS. It will blow your speakers but other than that, it sounds oddly pleasant. You go find it, I won't be responsible. In the normal world however, LUFS measurements are always negative. For TV broadcasters, the aim is -23 LUFS in the UK and Europe and -24 LUFS in the USA. The USA is quieter than Europe - that's a first.

Along with broadcast recommendations for LUFS, we also have recommendations for peak levels. In this case, true peak. Instead of dBFS, full scale, we have dBTP, true peak. In digital audio we have sample peak. So the highest any individual sample can be is 0 dBFS, zero decibels relative to full scale. But we can also have inter sample peaks. This is where a signal at the waveform level during recording is on an upward trajectory and hits 0 dBFS just before the next sample is measured, then goes through zero again on the way down. In the audio that's recorded, the highest sample value is 0 dBFS, but the analogue signal went higher between samples.

On digital to analogue conversion, this will result in an inter sample peak, where the level goes over 0 dBFS. A competently designed converter should handle this gracefully, but a lesser converter may produce an inter sample clip. This can also happen in sample rate conversion. I have to say that if I've ever heard an inter sample clip, then it was mixed among a whole load of other distortion and its difference in nature didn't register, but I prefer clean audio and the risk of inter sample clips is unacceptable. Broadcasters think so too, so alongside a measurement in LUFS it's good also to account for true peaks, not just sample peaks. The requirement for TV broadcasters is that there should be no peak higher than -1 dBTP in Europe and -2 dBTP in the USA. This keeps things safe and clean.

We can apply LUFS and DBTP to music too. Streaming services often provide the ability to control loudness. I say often - it could be all of them, but there are far too many services to check and some services enforce it. This can be done by bringing down audio that's too high in level or bringing down audio that's high and raising up audio that's low, on a track by track or whole album basis. What's best is to distribute your music at the right level, so that streaming services don't have to mess about with it. Different streaming services have different requirements. Things change from time to time but the current information I have is that YouTube and Spotify prefer -14 LUFS, Apple Music -16. To achieve these rather high levels you may have to use a brick wall limiter and the more you limit, the more your sound is degraded. I compromise at -16 for my YouTube videos. I'd prefer -23 but then I'd be quieter than everyone else and that wouldn't be good.

It's also wise to limit your work to -1 dBTP. Now to do this, you will need a meter that can work with LUFS and a limiter that can handle true peak. There are plug-ins that can do the whole lot, analyse your audio, then set the LUFS level and limit it to whatever you need. I very much recommend you get one. One anomaly I've come across however, is that on Spotify, volume normalisation is on by default, so you hear tracks at consistent levels. If I analyse this by highly secret methods, I find that the level is -14 LUFS, just what I expect, but if I switch volume normalisation off, I find that many recent tracks are way above this level, approaching -7 or -6, which is crazy loud.

This weird logic has it that a songwriter creates a beautiful melody and lyrics, a singer performs it from the heart, backed by musicians who've made it their life's work to perfect their craft, recorded in a studio that possibly costs up to a thousand pounds a day or more, mixed by a specialist mix engineer with ears made from finest gold, then the label boss says, master it as loud as you can. Really, it does nothing than degrade the sound and since most people won't have a clue that they can switch volume normalisation off, it will be lowered in anyway. So all that happens is that it sounds bad. I have no answer as to why they do this, all I can do is suggest you ponder whether you should do the right thing or do what other successful acts are doing.

Hey, do you know what? That's it, all you need to know about game staging. I said at the start of this podcast series that it isn't complicated and it's hard to go wrong, but under the surface there is actually a lot to think about. If you want to cross all the I's and dot all the T's. I'm David Mellor, Course Director of Audio Masterclass. Thank-you for listening.

Thank-you for listening and be sure to check out the show notes page for this episode where you'll find further information along with web links and details of all the other episodes. On, and just before you go, let me point you to the soundonsound.com/podcast website page where you can explore what's playing on our other channels.