Headroom Questions

Question

Hey,
I've been reading a lot of discussions about headroom, and I'm hoping someone can clarify a couple things that are confusing me.
It seems the common wisdom is to leave about 6 db of headroom for peaks and 12 - 18 db for average signal level.
Here are my questions:
1. Is the -6 db peak level in reference to a recorded audio file? In other words, if I put an audio file on a track in a Logic Project, and insert only a Level Meter on the channel strip, and look at the Peak and RMS levels of the Level Meter, ideally the Peak should be roughly - 6 db and the RMS signal should be -12 to -18?
2. Assuming the answer to #1 is yes: I don't quite understand this, in that, if the peaks of the audio file are 6 db below the max level, what is the headroom for? "inter-sample" peaks? I realize this may be a dumb question, if so, sorry!
3. Assuming again that the answer to #1 is yes: Does this apply to ANY audio file, including a pre-mastered stereo mix of a song? Because, I thought that when you do your final (pre-mastering) mix, you should try to get Output 1-2 level meter to be as close to 0 as possible, which I found results in the peaks in the mix's audio file being close to zero. Am I missing something?
Thanks.

Answer 1

This is a very complex subject but basically resolves around how sample value(s) are handled inside a computer.
One crucial thing to remember is that when you are looking at a meter in Logic (or any DAW) you are not seeing a level, you are seeing a sample value which is representing a level. It has nothing whatsoever to do with volume.
Once a signal passes through your A/D and is inside your computer it is nothing but a string of 1's and 0's. There is no sound inside your computer, there is only maths. A single track by itself really presents no problem but it is rare that a song you are recording is going to be a single track. As you add tracks these sample values get added together, you may have come across the term 'summing'. Summing is exactly what it says, it adds these numbers together.
The internal mix engine in Logic will give you phenomenal headroom but we must remember that this needs to come back out into the real world, out of our D/A so we can hear it and play it. Very, very simplistically, what a computer does when a number is too big it will truncate it. Keeping your tracking levels lower (the -6db for peaks is cool, at 24bit it isn't harmful to take that to -12) will give the summing engine room so that nothing is truncated when coming back into the real world.
I like this question as it is an engineering question. When I started with tape we had to learn how to align the machines, de-mag the heads, clean the tape path, what tape had what tolerances and how hard we could hit certain tape brands to get a good sound. Now things have moved on but there is a new set of rules to learn. This one is the most crucial, I find.
You can experiment with levels, as I and many others have done and keeping tracking levels lower really does give a better sounding mix. As I said, this is a very simplistic explanation and you have to remember that plug-ins can clip too. Everything that you put onto a channel will 'add' to the numbers that the summing engine will need to process.
So, in your last question, a single track may not be clipping but 24 tracks all hitting -1, once summed with all manner of processing, will be way too hot.
Once you are finished and end up with a stereo song, you basically have 2 tracks in a single phase-locked stereo file. A mastering engineer will bring this 2 track recording up to a healthy level.
I hope this helps a little.