Hi,
 
I was wondering, what's the preferred mastering approach using a DAW?  Do people who master their own songs do so by placing mastering tools (plugs) on the main bus output or instead on the sub-mix buses (such as drums, bass, guitar, vox) prior to the final output bus?
 
Thanks,
 
        Sonic 
 
 

Export your mix first as a stereo wav. Then import that into a new project and add your mastering chain, dither and export. That's the best way.

35mm
Export your mix first as a stereo wav. Then import that into a new project and add your mastering chain, dither and export. That's the best way.

That's similar to the classic analog methodology, so to speak.  Is that approach still necessary now days with DAW's ? .... What would be the advantage of mixing out the song when you could simply put the tools on the final output bus inside the project?  

I think the general approach is to first make the mix as good as possible in itself, then at the mastering stage you focus on doing other overall adjustments. The issue is more from a perspective of making a set of songs consistent with each other, so that you can have en EP or album play song by song without radical changes in loudness, sonic image etc. To make it uniform as a collection of songs. If you make to many "mastering" type of decisions in the mix, it is harder to work in mastering modus in the aftermath. I think the main challenge with the mastering process it to make songs sound uniform sonically and for them to sound decent on any sound system ... from your 10$ phone to your 10000$ HiFi system. 

What is the purpose of your master?  Meaning, to what end is the finished product intended.  If its to hand around or enjoy for yourself, or as a test to help understand how the current mix might relate to the fine touch of mastering,  then the master buss works fine in the project.   I used to wonder what a mastering engineer would think of my mix so it was important to me in my learning curve.  I had to learn to let go and commit - to learn everything I can, and to continue to learn. 
 
If its to build a product (song) with other product (songs) for distribution - then mastering takes on a whole different perspective.  Creating a full CD length work of art where each song relates to one another - that takes a different area of expertise.
 
The rule that works for me - if its for profit, I send the work out to a qualified engineer.  If its anything else, I always export.  Before I import and start the 'polish' process (my name for it because I am NO mastering engineer), I give my ears and mind a rest.  For me, mixing and 'polishing' are two very different processes.  I go from hands on, say 40 tracks with fader, eq, sound field placement, sweetening and automation to 1 stereo track, where the focus is on the frequency range in the sound field, and then maybe a smidgen of limiting.  I find if I am hopping around with the other 40 tracks to make adjustments, its impossible to focus on the final output.
 
Thats me - YMMV..  Take care. 
 
 
 

Another approach with a surprisingly wide-held support is to actually mix right "into" your mastering chain from the very outset.
 
Admittedly  there are probably as many critics of this method as there are supporters - but it's worth trying out

Part of traditional mastering is assembling songs into a collection, as for an album. For that you need to work with all the songs as a group, comparing their levels and EQ so they sound like they belong together, as well as sequencing, crossfades, inter-track timing and placing indexes for track selection on a CD. In this scenario, it makes more sense to export individual songs with lots of headroom and then bring up their levels in a separate mastering project.
 
However, nowadays we're more likely to be simply posting a song on a filesharing site or YouTube as a standalone file. In that case, we're really only concerned with making sure our levels are appropriate for the target platform. You don't need a separate mastering step for that and can just do it all within the project. All you have to do is settle on a loudness/peak standard and stick to it for every project. Which standard you choose will depend on the target platform, but -14 LUFS and -1 dBTP works for most online distribution.

bitflipper's analysis was right on the money. 
 
The only thing I would add is to do the final mastering with "fresh ears".  After doing parts and tweaking to get the mix you want to export, it is always good to let it sit or a few days if you have the luxury of time.  Now, when you listen to that stereo product only, you will get a fresh perspective of where you want to be and, of course, where you are.   My own process has been to do a final mix in Cakewalk by BandLab and export that stereo mix to a designated folder.  From there I access it in Sound Forge which has been my preference mainly because I have used it from day one and my mastering plugins from Cakewalk can easily be ported over there for use.  (I put those plug-ins in a separate designated folder called Mastering in Sound Forge).  I sometimes will go the extra step of porting a mix into an app like Lurrsen or something like that...that is, of course, if I am not sending it out to a mastering house.  If you are indeed sending it to a mastering house, make sure you keep compression-limiting at a minimum and check with their requirements before sending it out.
 
I do suggest, however, that you read as much as you can about mastering because in the end only you truly understand the concept that you are aiming for with a song(s).  There is so much out there now, why not learn as much as you can.  Bottomline:  Take a few of your favorite pro-released songs and use them as "ear references".
 
While it is indeed true that a good mastering engineer is a gem to be appreciated, it does not mean that you can't learn and who knows, maybe you will someday be that mastering engineer.
 

35mm
Export your mix first as a stereo wav. Then import that into a new project and add your mastering chain, dither and export. That's the best way.

I do this as well.  It helps to separate the two processes (mixing and mastering) and helps me focus only on one job at a time.  It specifically makes me really get a perfect mix BEFORE going into mastering.

I always have and always will :
Write/compose= anything that can capture audio ( Cakewalk/Tascam DR40/Cell phone) 
Track/Record/overdub/edit = Multi track recorder/ DAW ( Cakewalk) 
Mix=Multi track/DAW  ( Cakewalk ) 
Master= 2 track/Wave editor  ( WaveLab ) 
Replication=Burn CD/ post MP3  ( Nero/ Gold Wave/  Internet ) 
 
I will always master in a Wave editor because that is what that software is made for. It is optimized to work with a stereo wave file. Workflow is superior ( for me) and all the proper tools are on hand. 
And as said if this is an album of songs it is much easier to produce a balance between songs using the tools.