Such a thing as too much headroom?

[Undrig]

Finishing up a track and it peaks at like -9db. I've heard conflicting reports that -10db is fine by some people and others say it should peak between -6db to -3db. While I realize I could redo the gain staging for the whole thing to raise it up a bit, is it really worth it? Shouldn't -9db be ok for a mastering engineer to work with? What disadvantages could I face as a result of leaving it so quiet? I think from now on I will check my kick and snare to ensure they peak around -6db to start and then gain stage the rest of the tune around it in the future.

[mromgwtf]

If you think it is too little then adjust your master fader and that's it. The only problem here is floating point precision, for example, a mix mixed down to -6db has two times less floating point precision than a mix mixed down to -0db.

[Undrig]

I've never really considered fiddling with the master fader to over/under compensate as a legit way to fix it. Otherwise the engineer could just as easily do that themselves and the concept of headroom would be negligible if someone made a track that clipped and simply turned down the master fader to fix it instead of actually fixing the problem at the source. I would assume redoing the gain staging (while a pain in the ass) would be more beneficial in the long run. At this point it's merely a question of whether -9db is too much headroom for an engineer.

[fragments]

lol, there is absolutely nothing wrong with -9db headroom. People say -6db to -3db because they don't really understand volume and gain staging. They think it sounds better loud more than likely.

In the digital realm there is no point in mixing really hot and as far as I know there is no such thing as too much headroom in the digital world as you don't have worry about a noise floor (or if you do the noise floor is so low its irrelevant).

Also, if individual tracks are clipping, turning down the master doesn't stop them from clipping.

If you think it is too little then adjust your master fader and that's it. The only problem here is floating point precision, for example, a mix mixed down to -6db has two times less floating point precision than a mix mixed down to -0db.

I wasn't aware of this. Why does that matter?

[Undrig]

Much appreciated. I am using some analogue kit but have already made the proper adjustments so they're not an issue with the mixdown. Thanks!

[fragments]

Much appreciated. I am using some analogue kit but have already made the proper adjustments so they're not an issue with the mixdown. Thanks!

Yea. Any signal you are running through external gear you'll want to make sure you get a nice hot recording with low noise. : ) I guess it doesn't make sense to have a mix peak at like -20dbs, there isn't any reason for it. But you could easily get that signal up to around -0db w/o any unwanted distortion or artifacts as far as I know.

I'm still curious about this floating point business though.

[GregoryTJ]

If you think it is too little then adjust your master fader and that's it. The only problem here is floating point precision, for example, a mix mixed down to -6db has two times less floating point precision than a mix mixed down to -0db.

I wasn't aware of this. Why does that matter?

Say you have a large size image (say 1000 x 1000), then say you resize it to 50 x 50 and once more back to 1000 x 1000, the image is going to be quite low quality after this.

Even though this is a horrible analogy, and they don't quite work the same, you get the idea I hope.
Also, if you render to an audio file and open the sound again, you have to take bit depth in to account too, which is just another type of resolution in the amplitude domain.

[fragments]

If you think it is too little then adjust your master fader and that's it. The only problem here is floating point precision, for example, a mix mixed down to -6db has two times less floating point precision than a mix mixed down to -0db.

I wasn't aware of this. Why does that matter?

Say you have a large size image (say 1000 x 1000), then say you resize it to 50 x 50 and once more back to 1000 x 1000, the image is going to be quite low quality after this.

Even though this is a horrible analogy, and they don't quite work the same, you get the idea I hope.
Also, if you render to an audio file and open the sound again, you have to take bit depth in to account too, which is just another type of resolution in the amplitude domain.

That's a good enough explanation for me. Thanks man. I can see how that would be an issue.

[Icetickle]

If you think it is too little then adjust your master fader and that's it. The only problem here is floating point precision, for example, a mix mixed down to -6db has two times less floating point precision than a mix mixed down to -0db.

I wasn't aware of this. Why does that matter?

Say you have a large size image (say 1000 x 1000), then say you resize it to 50 x 50 and once more back to 1000 x 1000, the image is going to be quite low quality after this.

Even though this is a horrible analogy, and they don't quite work the same, you get the idea I hope.
Also, if you render to an audio file and open the sound again, you have to take bit depth in to account too, which is just another type of resolution in the amplitude domain.

That's a good enough explanation for me. Thanks man. I can see how that would be an issue.

So if it's really quiet and you bring it up back to 0dB, the quality is gonna be lower? That is what i thought but my friend was trying to prove me wrong...

[SunkLo]

Headroom? No. There is such a thing as too much noise floor though.

In your final master you're not gonna want a huge gap between your highest peak and 0dB because you won't be maximizing your bit depth. Every bit gives you 2 times more resolution. 16 bit is 2^16, 24 bit is 2^24. A bit corresponds to 6dB. So normalizing a 16 bit wav to -6dB is basically like truncating it to 15 bit. Don't think that more bits mean more resolution and less jagged edges on your waveform. The actual signal output through a D/A converter is perfectly continuous like the original waveform. The only real difference bit depth makes is noise floor.

When a signal is sampled, it needs to be quantized to the nearest bit. For simplicity's sake, let's say we're recording 2 bit audio. 2^2 = 4 possible values for the waveform. Let's ignore the negative portion of the signal for now, we'll pretend like we have an analog signal that ranges from 0 - 1. Since we're recording 2 bit audio and have 4 possible values that the waveform can be stored as (in binary: 00, 01, 10, 11) Distributing those 4 values over a range of 0 - 1 leaves us with roughly 0, 0.33, 0.66, and 1. Say the A/D converter samples the analog waveform and gets a value of 0.5. The audio then has to be quantized to the nearest value so it can be stored in 2 bits. So the converter stores that sample as 0.66 or "10" in binary. This random discrepancy between the sample's value (0.66) and the actual original value of the waveform (0.5) produces distortion that manifests as evenly distributed white noise. This is called quantization error. So a higher bit depth recording will have more resolution, and as such the sampled values won't be so far from their original values. Meaning the quantization errors are lower; The noise floor is quieter.

In a DAW's 32 bit floating point internal resolution, you have excessive headroom so you really don't have to worry or think about it until you render out to a wav. You can clip the dick out of all your channels and then cut the master fader down, render to 16 bit wav, and it'll sound identical to a version that was gain staged properly. This is the beautiful flexibility of digital. However, plugins are where you need to be careful. Clipping the input of a plugin is asking for trouble. So gain staging is still important. It's also incredibly easy to mix when all your levels are consistent. Your faders correspond to the actual track's level, you can set thresholds intelligently, it's easier to assess the dynamics of a track, etc. It's also crucial in maintaining a consistent monitoring level. Otherwise the volume starts to creep up as you add more and more; things sound better and better; you think you're on the right path. Then you compare it to an earlier render, matched to the same volume, and realize you've muddied up your mix.

When sending tracks for mastering your main concern is "Is this mixed as cleanly as possible?" If your master's clipping, the answer's probably no. But if you've got a clean mix and it ends up coming up a bit short of your target, you can just boost the gain. Since you're still in the realm of 32 bit floating point, you can scale it up or down with pretty much no consequences. This doesn't hold true if you render to 16 bit wav and then try to turn it up or down. As long as it's in a sensible spot your mastering eng won't mind.

You should note though, sometimes people use headroom to refer to the difference between the RMS of a signal and its peak, in other words the crest. This is much more important than the headroom between your highest peak and 0dB (assuming you're not clipping) There's dynamic meters out there that will tell you your peak and average values and the dynamic range between them. Brainworx' bx_meter is a nice one, I'm pretty sure there's freeware meters that do the same. Throw one on your master and look at the dynamic range of the signal. If it's telling you your track is too compressed, there's nothing your mastering engineer will be able to do. Doesn't matter if you turn the fader down to -100dB; If the dynamics of the actual signal are already squished, your mastering job will be crippled.

[Sharmaji]

if you're working at 24 bit, which you should be because it's not 1998, there's no such thing as too much headroom.

These days when working up mixes for clients, once i get my kick worked into the mix, i'm generally peaking at around -18dB. That way, when everyone involved wants X instrument louder, i have enough room to work with whatever level, dynamics, parallel, etc to make mix changes happen.

If you've mixed to -3dB, everything sounds great, and you realize after a break that you need more kick and more lead vocal/synth/etc-- what'cha gonna do? you don't have the space for it. Or if you realize that the mix really needs some EQ and compression on the master-- where are you going to find the room for it?

@SunkLo the waves version of the durroughs meters are the best i've come across for really letting you know how much power is in a mix.

[SunkLo]

@SunkLo the waves version of the durroughs meters are the best i've come across for really letting you know how much power is in a mix.

Oh god yes and they're gorgeous. I want to buy a bunch of second hand LCDs so I can have a bridge of them above my main monitors.