I'm just getting started with mastering myself.  Check out this tutorial on Youtube.  Very simple, concise video that will start to demystify the process for you.  I don't doubt that I have a good few years to really learn how to master well, but watching these lessons really brought me up to speed in regards to getting a final project with a respectable volume level.

Check the videos here:

http://www.youtube.com/watch?v=_S6s4Sr6YJ0

A great multi-band compressor that you can download for free is T-SLEDGE.  I picked it up month or so ago and it has quickly become a staple plug-in in my work flow.  A quick Google search should direct you to the site.

Part of my mixing process is to put a little compression and EQ on the Master bus (you'll see what I'm talking about when you watch the video).  I then get my Master level up to around -4dB to -6dB.  Once I have all of that, I export all of the drums on one mix, all of the string instruments on another mix, all of the vocals on yet another mix.  I then open up a new session and fly them all in on their own tracks.  This is the long way to do it.  You can also simply have an instrument track and a vocal track.

Once you have all of your stems flown into the new Master session, add another compressor and EQ onto the Master bus.  Automate the volume on the vocal track as necessary.  Push the Master fader as high as possible without ACTUALLY HEARING any clipping.  Disregard what the meter is telling your eyes.  Listen with and trust your ears.  If the meter says you're clipping, but you don't HEAR any clipping, then you're fine.

Once you're satisfied with how all of that sounds, export the whole thing as a final mix.  The last thing I like to do is check the mix levels on as many different meters as possible.  Fly the mix into Audacity and hit play.  If you've done everything right up to this point, you should be maxing out at zero.  If your computer has a sound meter, go ahead and check what it's peaking at too.  Again, if you've done everything right up this point, you should be peaking at zero.

If for some reason the levels still aren't as loud as you'd like them to be, select Effect from the Audacity tool bar and select Compressor.  Usually a 2:1 ratio is best, with most of the levels on the other controls turned way down.  Remember, you've already done a lot of compressing at this point and you don't want to cause any pumping.  

Once the compression is applied, you should have a nice big fat wave form, a nice thick sound coming out of your monitors, AND a dynamic range that is distinct in its loudness and softness (snare drum popping, electric guitar screaming, bass guitar thumping).  Export as the highest quality mp3 (you can crank up the quality by clicking Options on the Export dialogue window) and you're all set.

Now, I'm sure that the way I'm doing all of this isn't the easiest or best way to achieve a master with good volume levels.  Just saying that this is what I've done and it's working great for me so far.

Cheers,
Reid

 I agree Kalle. I believe it was already covered in the other posts above that to trust only your ears in a digital recording can prove troublesome. I was amazed at the clipping that was happening that I never heard. The result of careful spectral analysis can clean up a recording. Just as a person can't believe everything they see,the same applies to hearing it. I have been tested to have good hearing(this surprises me lol) and it is all too easy to miss the clipping.

  Yes I totally agree that we really don't know what natural sound is anymore. Almost everything we hear is compressed to the hilt in the digital domain. 

  I am working on another mix right now that is nothing more than piano with dynamics and vocals with a few pads thrown in. Believe it or not,that mix is a challenge. I finally got the mix under control at around -3db. I had to stage compress to achieve that result. The interesting thing is that according to my RMS readings I am still at around -12 db in "perceived loudness". In order to bring the mix up to a more suitable -6 to -8 RMS I need to further treat the mix.

 One thing that has also helped  is laying my monitors on their sides so that the tweeters line up with my ears and since I use ARC, I set it up for several  scenarios like door of studio closed,door open etc.

I would guess thsecon one is the Alison Krauss, but I am probably wrong.

Her new album, Paper Airplane has some decent dynamic range on the title song.

 Am I looking at the graphic wrong or is there still clipping on those tracks?

A bit to the side, but very efficient and effective:
 
I use a MAIN OUT bus that controls the monitoring volume sent to the audio device.
 
The MAIN OUT bus does NOT have "effects." 
 
The MAIN OUT bus has MONITORING tools: Voxengo Span, vintage meter, Panipulator.
 
The MASTER MIX bus is the master bus for Sonar's output.
 
The MASTER MIX bus is used for bounces. All final effects are in the bin.  This bus is output to the MAIN OUT bus.
 
I have at least two blank tracks, labeled MIX1 and MIX2, in the project template.
 
Before bouncing to tracks, I change the name of the mix tracks (add the date) so the wav file will have a reasonable name.
These MIX tracks output to the MAIN OUT.   Therefore, no need to bypass the effects/eq, or change the volume setting, of the MASTER MIX bus when listening to a mix track.
 
I then solo the mix track to listen to the mix(s).
 
And because the MAIN OUT bus volume is the same, I can compare, audibly and visually, the mix with original tracks.
 
Per CJ's insight, if the mix does not sound correct, before I do anything else, I use control-z (undo) to remove the mix clip, and the wav file from the disk. This means I have fewer orphaned audio files.
 
When a mix is happy, I copy the mix wave file to a folder in the CD projects folder for CD creation. I use Sony CD Architect. This program reads 32 bit files and applies dither during CD making.
 
This way I do not have to worry about dither until it matters. I do not have to remember, or make note of, which files have been dithered/made to be 16 bits. All mix files 32bit.
 
I seldom use the export function of Sonar, except when directly needing a 16 bit file.
 
J
 

ALWAYS create a restore point before installing ANY hardware driver.