Your friend is sort of right. All sound levels can be measure in "decibels", which start at 0db. However, the meters on your audio equipment are normalized to 0db, which represents the highest signal that the system can supposedly handle.
In the digital realm it's a no brainer. There is no value over 1111111111111111 (unsigned, don't jump on me computer geeks), so that value is 0db. Anything "over" 0db in the digital realm is 0db, because there is nothing over 0db
A digital meter always stops at 0db, so it looks like -6db -3db 0db.
In the analog realm it's a bit trickier. 0db represents some maximum voltage that the equipment works with...but we all know that no two analog systems are the same, so some will have a little bit of room over the 0db mark on the meter, some may have less. You can tell an analog meter very easily because it will look like -6db -3db 0db +1db +2db (in other words, there are values ABOVE 0).
Think about your old tape recorder...when you set it up to record, you'd set the record level such that the meter would bounce in the red every so often. It wasn't a big deal because tape by nature offers a little bit of "extra" peak room without sounding bad, and you want your level as hot as possible because the noise floor is so high with tape. (taken to the extreme, this is referred to as "tape compression") In a digital recorder, on the other hand, you want to set your level such that the meters NEVER hits the red line....all samples that are over 0db are truncated down to 0db, and what you end up with is very nasty sounding distortion.
As far as where you need to track when using a digital system, it does kind of depend on the quality of that system (how good the converters are and where the noise floor is), but in most cases it's like this: in a 16bit system, you'll want to track such that you're peaking at around -3db or higher. In a 24bit system, you'll want to track such that you're peaking around -12 to -6db.
The reason you have so much more "room" on a 24bit system is that a) the noise floor in a 24bit system is often lower and b) you've got a lot more resolution to work with.
While 'a' above is pretty easy to understand, 'b' is kind of tricky. The decible scale is logarithmic, and without going into too much detail, you lose about one significant bit of resolution for every 6db of level reduction. So, in a 16bit system, there are 2^16 - 2^15 = 32768 values representing the range -6db to 0db. In a 24bit system, there are 2^24 - 2^23 = 8,388,608 values representing the range -6db to 0db! Therefore, you can *theoretically* track 48db lower in a 24bit system and still have the same resolution as a 16bit system. I say theoretically, because you do always have to contend with that noise floor...and tracking at -48db would require a lot of volume adjustment at mix time which won't be pretty
Slackmaster 2000
