Best Answer:
"In 24bit recording, anywhere from -24dB to -12dB is GREAT!!! This gives you 12dB for any occasional spikes and this assures you'll never clip and waste a good recording take. If you think you need to record as hot as possible to get quality, your (sic) wrong."
its "you're" CJ, a contraction of " you + are"; but none of us expect you to get it right :-)
Wrong Answer:
"There is some headroom built into 0 dB"
??? I don't think so.
Most Confusing Answer:
"I also tend to write out the song's sections, in order of importance, and mix to THAT list's order, which usually them puts all the hoopla and bells and whistles and such of the end of the song mixed first, and then work backwards still in order of importance, pulling things back and even muting some stuff along the way, so that it gets easier and easier as you follow things in this direction, pulling a piano part back or dropping some steel string guitar for the second verse, and maybe then being able to leave in some neat flute or french horn sound that otherwise would be too much.
 
It just makes sense to me that if the end of the song chorus has to be the maximum presence version of the chorus, then starting with it and doing that - giving it everything you can, gives you a known point to keep the prior sections down below, and getting things to calm down is a lot easier than going the other way.
 
Doing things this way often saves a lot of time, as you may now realize that you might not actually NEED that 10-guitar multi-track 'thing' in the 2nd verse anymore, because it just doesn't fit within what you already KNOW works for the climatic ending choruses and such, because working backwards from the maximum doesn't leave room for that number of instruments without causing massive problems,"
 
What does this have to do with "tracking levels"?
 
Calibrate your equipment, 0db on a vu meter should be about -18 to -16 on Sonar's meters.
As stated by many above, "peaks" at -6 won't hurt anything. Actually any "peak" that does not exceed 0 won't hurt anything. But I do believe that 0 represents the absolute maximum (as in we are now out of "headroom").
 
Tom
 
ps: CJ's advise was/is just as applicable to 16 bit recordings as well. Trying to get "as hot as possible without going over" is/was always a fool's errand. This requires that your outboard gear be running at max gain. Most gear tends to get "edgy" at this point. Think Plexi Marshall with all controls on "10". Great for guitar, not so good for most other applications.
Personal experience;
my early digital recordings all "sucked" for this very reason.
Too much gain on the Alesis 3630 going in.
"Grainy" as hell
Just sayin'...

Tom,
 
This is great, but (and yes, I have been reading posts about how you pro's have tools to measure this and that, and I could not be more lost, although I kinda/sorta get the jist of it and DO see the importance, I just don't know where to start) how and what do I use/do to calibrate anything.
 
I just thought that if my interface was never in the red and Sonar's meter for any track did not go above -6 then all was good.
 
If anyone wants to give me a bulleted list of what to do to have calibration and a measurement tool for the instances I need this or these for, (please define what the intent is and the method/tool required) I would be so very grateful. Or, maybe there is a link or two on this.
 
Let's face it...you pro's are light years ahead of me.
 
I read this forum every day and copy and paste stuff to a document, but I can only keep up with so much at one time and yes, there is a LOT to learn. I am reading and watching videos, but heck I am not even sure I know the difference between the dB measurement and the rms measurement. And don't get me going on how to use a compressor like you pros do. I am working on understanding compressors and think I have a handle on what the overall end result or hi-level view is, but to know what tool to pick and why... and of course knowing how to use it like you pro's...well I am still working on this.
 
I swear...if I had time an money I would seriously be going to engineering (sound) classes at night.
 
Thanks to all of you for all of your help.
 

1. You don't ever want a stray peak to go over 0dBFS. Ever. So leave ample room. And then some.
 
2. You don't want noise from your audio interface to be audible, especially if you have to add gain to the track later. But, especially with 24bit audio, this is not likely to be much of a problem with a decent interface, so you can pretty much ignore it.
 
3. Therefore see #1.

Trem,
Trust me I am NOT an audio "Expert".
That being said:
In re-reading your question(s), you are "fine" with the MOTU gear.
If you're peaking at -6 in Sonar, all is good with your signal chain.
I also see why Robert went to such lengths in explaining "good" mixing practices
(Robert my sincere apologies for the label "Most Confusing").
Carry On!
 
T
 

A few fine points...if you record at -24 dB, you're losing about 4 bits of resolution. With 24-bit converters, due to quantization noise, circuit board layouts, "glue" components used, etc. you'll often have 20-22 "real" bits of resolution. So taking off 4 bits now yields 16-18 bits of resolution. However - and this is important - those are 16-18 real bits, in other words, even at 16 bits you have the full dynamic range of 96 dB. With 16-bit converters, you don't get 16 "real" bits but more like 14 and in poorly-designed consumer gear, sometimes as little as 12.
 
I've found that if I aim for peaks to hit -6, then I'm not going to go over. But I have pretty predictable levels. The advice about leaving more headroom can save a session. I'd much rather have an increase in noise that's for all practical purposes theoretical than an "over" that ruins a take. -12 should be enough to accommodate over-enthusiastic vocalists or whatever; it all depends on how safe you want to be. With a live recording where you only get one shot, -20 or even -24 may make the difference between having a recording you can use and one that's useless.
 
The only other potential issue with recording at a low level is needing to bring up the track gain control to hit compressors and distortion hard enough when mixing. However, with SONAR's available 18 dB of gain, this shouldn't be a problem. You can also increase the level of the audio itself once you have audio that's been recorded without distortion.
 
Finally, there's a phenomenon called inter-sample distortion. Most meters indicate the value of the samples that are taken, but not the result of those samples going through a smoothing filter on playback. The smoothing filter can actually create levels higher than those indicated by the samples. So it's also good practice to leave some headroom on playback at the master out in anticipation of the audio going through a smoothing filter after being converted back to analog. I've found -3 dB is usually enough, but not always.
 
SSL offers a free download on their site of a meter plug-in that can indicate inter-sample distortion. It's worth inserting into your master bus so you can make sure your levels won't lead to distortion on playback.

TremoJem
I am working on understanding compressors and think I have a handle on what the overall end result or hi-level view is, but to know what tool to pick and why... and of course knowing how to use it like you pro's...well I am still working on this.

You might find this article helpful.
 
The following may sound facetious, but I'm sincere...all that matters is whether the music has emotional impact for a listener. If technical limitations prevent that from happening, then zero in on one limitation at a time and fix it. For example if the limitation is "sounds too bassy," time to learn about EQ. 

Anderton
A few fine points...if you record at -24 dB, you're losing about 4 bits of resolution. With 24-bit converters, due to quantization noise, circuit board layouts, "glue" components used, etc. you'll often have 20-22 "real" bits of resolution. So taking off 4 bits now yields 16-18 bits of resolution. However - and this is important - those are 16-18 real bits, in other words, even at 16 bits you have the full dynamic range of 96 dB. With 16-bit converters, you don't get 16 "real" bits but more like 14 and in poorly-designed consumer gear, sometimes as little as 12.

 
It should be added here that once the quantization error level is more than a little below the noise from the room/mic/preamps/etc. you will gain nothing by having more bits. Those other noise levels will ultimately determine how many bits you need for recording.
 

 SSL offers a free download on their site of a meter plug-in that can indicate inter-sample distortion. It's worth inserting into your master bus so you can make sure your levels won't lead to distortion on playback.

Realistically, if you leave at least 3dB of headroom you will never get distortion due to intersample peaks for anything except test signals.

One thing I have noticed is that the pro guys who started sharing the idea that you don't have to record with 24bit as hot as you would in 16bit seem to have pro grade A to D converters where 0dBFS is calibrated to +24dBu at the analog input. Their preamps are running in a sweet spot and they can bring stuff in at +10dBu and leave lots of analog and digital headroom.
 
This scenario is a lot different than using a cheezy usb powered I/O box where the 0dBFS on the 24bit digital side is calibrated to +6dBu at the analog input. If you have a preamp that runs easily up to +20dBu but you can't use it past +6dBu... you may never be in a sweet spot.

I've never met anyone who can actually hear an intersample peak during music playback.
 
If you are running at 44.1kHz an intersample peak will be a very fast chirp happening at 22.05kHz or greater. Good luck trying to hear that.
 

Love the link Craig.
 
Love the chattering back and forth about the dbu etc.
 
Although 50% of it is over my head...I am working on figuring out the rest.
 
Hey, that is what Google is for...and of course, this forum.