It was more a 16 bit thing, does not apply now with 24 bit resolutions. It is silly to record at too high a level . You won't gain a thing other than the possibilty of clipping the signal because you are too close to 0 dB FS!
 
Choose  K system ref level and work there.

Bristol_Jonesey
There is a thoroughly decent article in this months SOS written by Matt Houghton on the whole subject of Gain Staging.
 
It's well worth a read if your current knowledge is a little sketchy.

Couldn't find it online. Was it just in the hard copy? Any links?
 
Thanks,

Short answer: If its analog, levels matter, it its digital, especially at 24 bit, not so much any more.    I just read a good article on this in Sound on Sound September (US issue pg 136).  This should be mandatory reading as its very informative.  Digital level and gain staging not longer need to be at the the highest you can get.   However, gain staging is critical when summing many tracks.   Analog equipment and levels however still are seriously impacted by the level and the sound will change depending how you drive that equipment.   Plugins that emulate analog equipment also depend on the levels but basic channel levels like within Sonar per channel do not and you do not need to push the levels that high.   

Jeff Evans
It was more a 16 bit thing, does not apply now with 24 bit resolutions. It is silly to record at too high a level . You won't gain a thing other than the possibilty of clipping the signal because you are too close to 0 dB FS!

Agreed...
Also, when summing (mixing) multiple tracks close to full code, you'll have to attenuate the level of each track.   

To add what everyone is saying, you also have to watch the gain staging into your FX. It's a bad idea to make a recording and normalize the recording so every channel uses every available bit. In this scenario, you may end up distorting as you run through FX that are modeled after real units. They also model the behavior when the signal is too hot.
 
:-)
 

brundlefly summed it up perfectly when he said "the digital sweet spot is pretty wide, and signal to noise ratio in the acoustic environment gets to be an issue long before you run out of digital resolution at 24 bits".
 
You can think of each bit as 6dB of dynamic resolution. Not using the topmost bit means decreasing the signal-to-noise ratio by 6dB (don't anybody nitpick over my numbers here, this is rule-of-thumb stuff). With 24-bit audio, that means you've raised your theoretical noise floor from -144dB to -138dB, which is still WAY below the threshold of audibility, and way below the analog noise you added yourself when you recorded your tracks.
 
To put that into perspective, your favorite tape-recorded records, including those that won Grammys for engineering, won't have a noise floor below -70dB. You can "lose" a lot of bits (10 or 11 of 'em easily) before you're as noisy as those Grammy-winning records!
 
One other thought...analogies to analog equipment don't work. It's true that every amplifier has a "sweet spot", but that mainly has to do with the (non)linearity of transistors and tubes, as well as staying sufficiently above the noise floor. Inside the computer, any non-linearity is an intentional effect, and the noise floor is so far down it's a non-issue.

your favorite tape-recorded records, including those that won Grammys for engineering, won't have a noise floor below -70dB

 
...and yer everyday 16 bit sample-size tops that by 26db...almost like the headroom on an analog console...
 

Some one mentioned the practical reality of acoustic noise floors.
 
Some one mentioned the great benefits of 32bit floating point math.
 
Those are the two important things to appreciate.
 
 
 
I like to record with full input resolution during the A to D conversion regardless of the acoustic noise floor. Then I can make full use of a noise reduction process in 32bit (or 64bit) floating point to make it sound less noisy.
 
That way I can savor the difference between POWr1 and POWr3 dither when I export to 128kbs .mp3
 
 
best regards,
mike
 
 
 
edit grammar 

 Mike, I discovered long ago that subtle humor is lost on this forum. If it weren't for the fact that I can count on you and drewfx1 to get the joke, I would never bother.
 
Anyhow, everybody knows UV-22 is vastly superior.

Upon close examination,
POWr algrorithm 1 and UV22HR produce very similar results