Let's talk head (room)

DM60

Well-known member
OK folks, doing some mixing looking at levels and I am noticing the peaks out of the master track. I am watching hits, and jabs from various instruments. No compression so we are clear on the master. Just stuff coming out of the channels.

I am not going to talk about dBs because I really don't understand half the stuff you guys are saying in that regard. So I will just talk, relative to clipping. Once again, numbers mean nothing in the digital world (I know enough about software and the value they show on the output is just some code dudes saying, "yea let's use this number for the output value for the visual and give it a value), but clipping is a pretty hard rule. Or at least most of us agree that when there is a clip in digital, it sounds like shit.

Now, had to qualify everything. Watching my meters because I want more cowbell. but it is getting lost, and I am already hitting red, so everything has to start coming down to get more cowbell. But if I don't start bring it down a lot, I will hit my ceiling for clipping probably as I continue to "tweak". Therefore I create a ceiling about 3/4 of the visual before I get red.

Now, I have mixed it and I like it. Ready to master (ok, for this discussion, for a collection of songs once known as LPs). If my final mix is too high, then it gives problems to mix it for a master/collection/what many of us know as an album (I never called them an LP) to get it all to gel and a unit? Therefore giving the final mix enough room between clipping and mastering so that it ca be worked into the final collection to make it all work together.

Now, here is my question. Do I understand the concept of headroom?
 
Pretty much. The term is really just a catchall phrase that can be applied to specific requirements as needed. In audio it started out as a way to gauge how much signal to get to tape or through some type of amplifying device in order to minimize noise/maximize signal to noise ratio. Now it seems to be most used as a general rule of thumb to allow enough dynamic range for mastering limiters to do their work. Among other things.
 
I would say DM60 that your grasp of headroom is, er, tenuous! (but then I think I am known here to be a techy, pedantic old fart?)

Headroom is perhaps best understood by looking at a mixer? Even cheap mixers these days run at an "Operating Level" of +4dBu (it's ok! Don't run!) or near enough one volt of signal. This 1V signal is usually indicated on the LED "vu" meters as "0" and that would match up with the vu meter on most decent open reel tape machines. But of course, audio signal levels jazz about all over the place and so the mixer needs to cope with that.

Firstly it needs to have low enough noise levels so that very quiet sounds are not lost in hiss. Any 1/2 decent mixer would have a noise output some 70dB below the OL. That is over 3,000 times LESS than +4dBu or well under one thousandth of a volt. Most are better than that. That is call the noise "floor".

But, of course, things get louder and again a cooking grade mixer would cope with +20dBu, ten times the OL. Good gear, 20 times. And that figure, above Operating Level is...Ta da! Headroom.

In the digital world it seems very different but it ain't really. Digital max is 0dB Full Scale. Explains itself really "top of the scale" and that is all yer bits so end of! the signal of course stretches down below 0dBfs and the noise floor for 16 bit recording is about -90dBfs that is 31,600 times LESS than 0dBfs. But we all run 24 bits? That puts the noise floor at easily -100dBfs or 100,000 times below zero. (top line AIs reach better than -120dBfs)

Perhaps you can now see the value of dBs? It gets bloody tiresome talking about thirty or hundred thousand divisions of signals!

The "mixer scale" is however restored for digital because we all record around -18dBfs don't we? Thus leaving a "headroom" of 18dB.

Of course, now that Home Recording kit has taken off so well we are bejaggered by all sorts of kit with all sorts of operating levels. Bloody minefield! But the basics inside the DAW still hold, record at an average of neg 18. People like Massive will thank you for it!

Hope that helps a bit? Techy, pedantic old fart out.

Dave.
 
Last edited by a moderator:
Dave, first I know you know your stuff and I would never doubt your knowledge. However, if we are strictly talking digital, headroom isn't as big a deal as it was in the analog world. If I mix something too close to 0 (clipping) and I want to master a collection, I just turn it down. In the analog world, since turning it down would introduce noticeable noise, then you would want some "room" to get everything to the same perceived level. I do think processing on the mix down if it is going to be "mastered" should be kept to a minimum (I would think mainly compression) so that the mastering process doesn't further squish the mix. But other than too muc compression, I don't see a big deal with headroom in regards of final mix.

Now for mixing (IMO), keeping some headroom helps in not having to go back and turn stuff down because you have pushed too far up and you have no room to work with. That way you keep the flow and you are not having to rework your mix all the time or use compression when you are not ready to use compression, just to keep it from clipping.

I really think the rules of analog keep influencing our thinking, but the rules are somewhat different. I am not saying I am right, but that is what I am thinking. Furthermore, when new people are reading this and the old gurus like yourself talk about this stuff, the references are difficult for many because for them, it is different.

But, I threw this out there to see what others thought and if my thoughts on the subject was really incorrect. Lots of guys on this board who know both worlds can "learn me" and others to improve our recording and mixing.

---------- Update ----------



That's what I was thinking when I said it :)
 
You are being diplomatic DM! I was really under the impression that you did not really know about " headroom, noise floors, operating levels, the works. You were confining yourself to the digital domain.

All I can say is I have come to understand that you track around -18dBFS because when you come to mix it is an additive process. Therefore the more tracks you add the closer you get to 0dBfs and eventually will have to pull everything down to "make room".

This "neg 18" idea is a hard one to get across to newbs sometimes? How many posts do we see asking "how can I get a high level in my DAW from my SM58?" Yes, it can be hard to get said noob to grasp that -18dB fs IS near enough 0vu on a mixer. They would not expect a mixer to hit +18dBu from a mic would they?

But then, as soon as you say the word "Mastering" I am outa here! All magic to me.

Dave.
 
To expand on that a little, as Dave says, headroom is a buffer zone of signal level above nominal operating levels to allow for normal variations in level (like transient peaks) before the onset of clipping or aliasing. Pushing signal levels beyond the available headroom results in distortion. So if you have a +4 dBu = -18 dBfs (line level) signal, this is an average or RMS value. It's easy to read on a VU meter with delayed ballistics. A slight curve ball here is how the different scales work. In the end, they all relate to line level somehow. With traditional VU meters, line level is usually calibrated to 0 dBVU.

A lot of the meters you find in DAW software will show you peak levels, rather than RMS or average levels. Some of them can be switched or can show you both at the same time. If you have a bar meter where the bar itself flutters rapidly but there's also a line drawn under the peak level that floats around slower, the bar itself is peak and the line is average or RMS. Line level should be read as RMS. Peaks can go higher but should not clip.

Transients (or transient peaks) are signals that are basically all attack and decay with little or no sustain. Steady state signals are the opposite, a lot of sustain with very little if any sudden spikes. An electric guitar slammed through a raging Marshall Plexi with the resultant distortion and compression is a steady state signal. Drums and percussion (like cowbell) are transient signals. Many types of signal are more complex and can have aspects of both transient and steady state properties. Acoustic guitar, piano and vocals.

If you have a mix with a bunch of steady state stuff going on, you can exceed line level without clipping. This is sort of how the Loudness War works. It happens at the expense of transient stuff that needs the headroom to balance well in the mix. Percussive stuff basically needs to be dialed back sufficiently to not clip while the steady state stuff is hogging all the sonic real estate. As you say, the answer is to keep everything low (or normal) enough that this doesn't happen.

You can get more cowbell that way.

It might seem counterintuitive because of the abused and hugely compressed levels of mastered audio from the past 3 decades or so, but it's a mastering thing. If your mix is significantly lower in level than commercially mastered audio, this is normal.
 
Even though numbers don't mean anything in the digital realm, we don't listened to digital. We listen to the signal after it is converted back to analog.

In analog, the closer you get to hard clipping, you run into soft clipping. This is why you record around 0dbVU (-18dbfs in digital), so you don't stress the analog circuitry going into the converters.

Once it is digital, it doesn't matter what level it is at (below clipping) until you bring the signal back out into the analog world again.
 
If you want a way to work without the maths and the numbers - look to your existing music collection for something in the style of your own music - Ideally a CD derived track rather than an isolated download. If the music is in iTunes or similar, just go to the folder with the files in and load each one in to a separate track on your DAW - hopefully one that gives you a maximum value the track reached. Play them all with the faders at unity and look for the quietest one start to end and the loudest one. Then see what the range is? Assuming these are properly mastered tracks, this should give you a decent ballpark figure for where your similar material should sit, depending if it's a loud raucous track or a quiet ballad one. As long as the thing never goes over it's really your call on where your version of loud sits compared to others - the important thing is that nobody playing one of your tracks needs to turn it up or down from somebody else's. The numbers you come up with work pretty well in practice.
 
If you want a way to work without the maths and the numbers - look to your existing music collection for something in the style of your own music - Ideally a CD derived track rather than an isolated download. If the music is in iTunes or similar, just go to the folder with the files in and load each one in to a separate track on your DAW - hopefully one that gives you a maximum value the track reached. Play them all with the faders at unity and look for the quietest one start to end and the loudest one. Then see what the range is? Assuming these are properly mastered tracks, this should give you a decent ballpark figure for where your similar material should sit, depending if it's a loud raucous track or a quiet ballad one. As long as the thing never goes over it's really your call on where your version of loud sits compared to others - the important thing is that nobody playing one of your tracks needs to turn it up or down from somebody else's. The numbers you come up with work pretty well in practice.

If I read correctly, this is what I was aiming at. And I should have been clear at the start, in digital, the numbers are a little more subjective. It really depends on how the programmer determines to display some digital value in the DAW. But, ever if that is true, by understanding what has been stated, summing values, peaks, RMS, etc. keeping headroom is important when mixing to allow the music to breath and have space. If the values are pushed to far up (near clipping), then one has to either go back and reduce tracks volume or start compressing. Compressing early in a mix just to keep from clipping starts to ruin the mix.

Therefore, for people trying to figure it, including me, headroom is important for mixing so the music has a place to breath. In digital, forget the numbers other than clipping, red is bad, OooKaaay. But, don't worry about loudness until it is in the final mix, either as a part of mastering, or as a final mix.

I think the idea of bringing in some commercial as a baseline for perceived loudness is a very good idea. Anytime you calibrate, you do it to some standard. This idea seems as good as any on a baseline for loudness.
 
...allow enough dynamic range for mastering limiters to do their work...
That's the important part in a modern DAW environment. Absolute levels are completely irrelevant until the very end when you're making your distribution masters. Up to that point all that matters is dynamic range. Any ME that can't just turn your mix file up or down to get it into their comfortable working range is wasting your time. You are rendering mixes to floating point files, no? ;)

But the dynamic range is actually much of what we do in mastering as well. We really do want the loudest peak on the record to be close to as loud as the distribution format can handle, and we work down from there to find an appropriate average level for each song or piece of the collection. If the DR of the mix is too big to fit in that window we've defined, we have all kinds of tools to squish it down or shave off edges or inflate the middle or whatever and (given the mix isn't just completely fucked to begin with) we can make it fit fairly well without doing severe damage.

IF OTOH, the mix has a lot less dynamic range than we've decided we have - when we put the average where we want it, the peaks are nowhere near zero, it might mean that it's just kind of a quiet sort of piece and that's ok, but it also might mean that it's been squashed a bit more than necessary, and could have bigger impact if the peaks actually did get up to the ceiling but something along the line stopped them short, and there is very little that we can do to a stereo mix file to fix that kind of thing without making an even bigger mess.

That's what people should mean when they're talking about leaving headroom for mastering in the digital world. If they tell you anything else, then they don't actually understand.

Now I'd be willing to bet that the actual problem with the cowbell is only a couplefew samples at the leading edge of the attack that blast way past all the rest of the thing. You could clip that off (preferably on the cowbell track) and solve the problem in the mix. In this case, I think that's more appropriate because anything done in mastering is ultimately going to affect the rest of the instruments to some extent. Honestly, too much dynamic range - when it comes from individual elements which are not themselves properly controlled - can be almost as hard to deal with gracefully in mastering as when there's not enough. There are more options, but it's still not ideal.

Also, the numbers you see on your meters are not even close to arbitrary. They are sort of relative, but they really do mean something important. -18dbFS means exactly the same thing in every DAW, every plugin, everywhere. They might do something different with that information, but it means the same thing.
 
Back
Top