I have used both Aaron's mp3 patch for Sonar and Audacity.  I would steer clear of the NCH stuff as you end up with shed loads of ad-ware for their other products in my experience. I usually export at 320 VBR and have had no issues with quality (44.1/16).
Individual track volumes should never exceed 0db in the digital world so -3db is a good max point.
In regards to overall volume as previously mentioned, if it is a final mastered version then peak at no more that -0.3 as a good level. If it is a mix (pre-master) have the overall output volume at between -6 and -3 db which gives some headroom to play with at the mastering stage.
Normally pre-mastering mixes are normally not for public consumption in any case so you would probably only export to mp3 if you want to hear the mix on a cheap reference system that doesn't support .wav files.
Best investment in my opinion is a good metering plug in or application which will give you RMS /Peak plus the increasingly important LUFS (loudness units) or optional K-system.
see Izotope Ozone as an example
 Top of my wish list for Sonar for mastering would be alternative metering (perhaps a pro channel module) which would show the master output in LUFS scale as it would save using a plug in or exporting to dedicated mastering software.
.
 
Hope this helps if only to give you a simplistic answer.

FWIW, as I'm only 6 mos into this shindig, I set my levels, mute any unwanted tracks, export audio to a wav file, open it in Audacity and convert to mp3. There are some cool tools to use in Audacity if you still need a little cleanup on the wav file before conversion. This is what I do during the production so I can get feedback.
 
PS - I put our first tune on Soundcloud. You can find it in the songs forum. It's called Son of the South and the drums are a tad hot, as we were changing the kit and checking to see if that's the one, which it is! So if that sounds crap, I need to go back to watching the videos again.

Look, the initial question was about how to prepare a mix that will ultimately be converted to MP3, and the answer is either:
1. MP3 is so transparent and true to the original mix that no special preparation will benefit the wave prior to conversion with a subtext of
  1a. the idiosyncrasies of the application you use to do the conversion is so critical to the MP3 fidelity that you have to choose the magic tool or your result will suffer . or
  1b. MP3 rapes your mix so badly that it doesn't matter what original mix you transcode it will all be equally bad. 
or 
  2. MP3 results in predictable (though possibly unknowable given the black box nature of the various transcoders) distortions of the original mix which are audible in the final result at least for some bitrates.
 
As a practical matter, those who so deprecate the whole process of lossy compression that they subscribe to 1b., can ignore the whole question. Those who subscribe to 1a can present evidence comparing the various transcoders, which has nothing to do with the OP's question. Those who subscribe to 1. are simply wrong, at least for low bitrate MP3. The OP's question implies that there is an answer that involves at least accepting proposition 2.
 
The issue becomes important in an era in which almost all music is delivered in lossy format, and some at quite low bit rates. Since most of the black box transcoders limit the user to selecting a bit rate, and the online services do not even do that, the logical way to manage the inherent distortion they produce in low rate MP3 is to alter the mix of the wave so that firstly the creation of objectionable compression artefacts are minimized and secondly important frequencies or spatial information is not lost. There are certainly people out there who are working on ways to change the mix for MP3 preprocessing. But if you try to find a detailed explanation of what that might entail you are probably going to be disappointed. If there were a simple answer, it could probably be encoded into a plugin, and if such plugs do not exist I expect they will at least be sold in the near future. Google mix for MP3, and you will probably do as well as you have here so far. It would be useful at least to have the capability of doing real time audio editing and have a low latency feed of on the fly converted MP3 into your monitors, so that you could use your ears without having to remember a sound before and after rendering. People who are mastering for the lossy world of the present, are going to have to train their ears, if not their intellect, to how their chosen transcoder screws with their original mix, and how they can compensate, as they did in the transition from vinyl to CD. Or they can just accept the results they get. 
 
http://www.justmastering.com/article-mixingformp3.php
 

slarta cuttin' to the bone! Heheh. I am most definitely approaching this from paradigm #2 (that lossy CAN result in degradation of quality and that degradation CAN be avoided if certain procedures are taken).
 
I'm certainly not saying that sometimes it is not a factor (like the ultra nice mix scenario) but that in the cases where it is a factor (like my congested dum dum mixes... which I'm trying to get better at) that I'd like to have strategies to avoid the degradation.
 
So it's one of those situations where to solve/avoid a problem you need to UNDERSTAND the problem which I don't (but thanks to all the great info here I'm starting to get my foot in the door on it all). In this case it seems, as I suspected, there is no "one size fits all" solution so it becomes even more important to understand the causes and possible remedies.
 
That becomes even more important in my case I think because my music (and how I like it to sound) is probably outside the "norm" of what most people are doing... AND even though I'm currently futzing with one particular style at this point (super layered/distorted/busy material with not a lot of room for the digital compression to work with... which is something I've learned as a result of this thread) I also want to/need to/will work with many other styles that may be more open/dynamic/easier to compress (like light ambient or acoustic stuff) or the exact opposite (approaching "Black MIDI" insanity).
 
And yes... this does all need to be done in a way that will play well in the new age of digital distribution. I'd love everyone to listen to my stuff at 24/48 at all times but that ain't gonna happen 95%+ of the time.
 
Someone earlier asked about plugs that can maybe duplicate the effects this type of digital compression as something gets mixed. If such a thing exists I would be interested (although I'm sure anything like that would be super expensive mastering tool stuff). Like toss the bugger on the Master bus in a mix project and be able to emulate the effects of compressing to mp3 at various qualities and other lossy formats so I can mix for ALL formats.
 
I have just recently put my VRM Box through the ringer on a recent mix testing my work at every stage to see how what I'm doing responds to the various room and speaker emus. I like that workflow because it exposes a lot of nastiness I could not hear before and I think it has really helped. If something like that were possible for various compressed file formats I could see it being really useful and could take a lot of guess work out of all this.
 
We've got bit reduction plugs (that are more for effects) so maybe the sample yanking properties of digital compression can be emulated. I'd imagine it would have to be a SUPER heavy duty lookahead tool but yeah... I'd like that.
 
Maybe this is the type of thing the Izotope (and similar tools) are capable of.
 
That's totally off track though. This thread has been awesome for giving me things to try and research... and yes... I DO overthink things. I don't view it as overthinking though. I view it as learning about the craft I love and making it so I can get sh*t done the way I WANT it done instead of struggling with simple crap.
 
Thanks again.

Not at all, Vast. That's quite helpful and I'll take a poke at that plug. Like I said I have (or had) no idea where to start researching. Now I'm cobbling together a muddled path forward. Before it was all googling "durrr... how maik mparTHREEVES sound teh gooderer... hurrr?!". Now I've got some reference points/keywords to start with from people who have actually had some luck with it all.
 
THanks.

Beepster
 
 
... Just hearing what others do and gathering up terms and concepts to research further is totally what I'm after.

 
Beep -
 
I went thru an exhaustive process trying all of the different rates and such for converting my CD collection to MP3s.  After trying a wide variety of styles with wide variety of dynamic ranges use of acoustic vs electric instruments, I came to the conclusion that 320 CBR was the best all-around to use.
 
So I did.
 
I have the same collection ripped to WAV so I can use whichever as the devices & circumstances allow.  I'm not sure what other front-ends support, but LAME from within Sonar offers those seen below.  I pick 320, Stereo and slide the quality slider as far left as I can.  I've never even tried the low & high pass filters.  Maybe somebody else can suggest when it might be appropriate to use them.