Recording Dynamic Vocals

Hi
I am recording a very dynamic song using Sonar X3 Producer.
The music tracks are all complete and I'm ready to lay the vocals.
The dynamics of the vocals are very ...well...dynamic.  There are parts where I almost whisper.
Other parts where I sing mid-range and other parts where it's all out.
I am using a Roland QuadCapture interface to connect the mic to the computer and within the interface are 2 preamps and a compressor.
Would you suggest I set the levels for each section differently for each section and record each section separately...
or set the levels for the whole song and use the compressor for the more dynamic parts.
Thanx

When I sing and record really dynamic stuff I always record on separate tracks for the "quiet" and the "loud".  This way you can independently set the compression and reverb and everything for the quite parts, then set them appropriately for the louder stuff etc.  Seems to work really well.  

You can use compression on the way in but I prefer to use it set for limiting. Very high threshold, fast attack etc. That way it won't do anything for most of the time but if any really loud bits happen it will just limit occasionally and stop the signal form hitting the red digitally.

I compress a little on the way in (in the pre-amp)
 
But if you have an extremely dynamic performance?
 
I would comp it... And sing the heavy parts from a bit more of a distance, and with a lower level.
 
And then on the quiet/whisper parts, just track them differently, boost the pre? And then use envelopes and compression/limiting to smooth it all out in the end. I sing quite a bit, so I'm not shooting from the hip. Most either can't or don't sing? Those that do struggle with it, those that don't work with those that do. Mic technique cannot be overstated, but it's not everything. If your performance is as you describe then you need a combination of, 1. Mic technique and pre-amp settings to tame the violent bits, 2. Some compression on the way in, 3. tools In The Box (ITB) once the comped tracks are there. You should be able to get what you're looking for if you give careful attention to each part of the process. 

This is why recording is an art, not science.  As the above posts show, there are plenty of ways to skin a dynamic vocalist.  You need to find the way that works for you, so start experimenting early.
 
vocalist who work the mic as they do live by backing up and getting closer to the mic can cause all kinds of havoc w/ the tone.  Cardiod mics have proximity effect, which results in more bottom in the vocal signal.  Also, if the vocalist gets close, small head movements can also change the tone.  That may not make a difference to you - but might.  Either way, it is something to be aware of.
 
That is one reason a good room, a good mic and good preamp help.  You can crank the input on the preamp up then and capture some of the more delicate stuff w/o crapping out the signal when it gets loud (and not have the room ruin the signal, too).  During mixing automation and compression can help level the signal.
 
I usually run vocals through a nice channel strip and keep everything on, even if I'm not Eqing (except a HP filter) and only compressing a few dBs just to firm the signal up a little, even if it is parellel compressing.  I'll usually start at a lower gain, just because many (if not most) musicians tend to play/sing louder during an actual take than during set up.  I can then add gain incrementally during takes.  If we take the time, I can even ride the levels a bit during the final take(s).  I'm a big believer in performance, and seldom comp takes except in sections.  Comping word by word seems to be the rage these days and is de rigueur in pop-oriented stuff, but seems like a way to up billing time to me.  Doesn't mean I won't do it if technically necessary, but I find it often distracts from the flow, perfection instead of authenticity.
 
@