(1) Working forwards from the tracks:
-- I direct the track output to a bus


(2) Working backwards from the master: 
 -- I feed the master only instrument busses (e.g. guitars bus, vocals bus, drums bus, etc.).

Sometimes the output of (1) and input of (2) are one and the same, but often I will have intermediary busses (such as a Tom bus, or an Acoustic Guitar bus and a separate Electric Guitar bus).  This is where I get creative depending on the needs of the project;  sometimes with compressors or eq's occurring in-between on those busses, or even reverb, etc.  I use Sends only very occasionally, but that's just me.

What I like about my setup for (2) is that I can easily generate a set of bus stems that are "final" components of the overall mix.  This means that (1) must always be in place for this to happen.  I'll set it up this way even on an easy project where the busses are somewhat redundant, because I find it easier to switch quickly from project to project knowing that this setup is in place no matter what.

John

Beepster


After seeing how routing is done in Pro Tools I am very happy with the way things work in Sonar. Sure there could be a little extra flexibility but the PT way just seems crazy/unnecessarily confusing to me. I like having a the very defined and obvious difference between tracks and busses in Sonar. Mind you I've never worked with analog aside from little cassette based multi trackers so I'm assuming PT is a little more similar to how things used to be in older studios.

I have used analog mixers and they are not very flexible. I prefer digital mixers and they also have flexibility problems. Most mixers have a set number of buses. I'm not too sure about the  remark about analog consoles from WDI. I suppose that the very expensive ones may have that ability. None of the ones one would use in a project studio or at home in general have that ability. Most mixers describe in their name the amount of tracks and buses they support. 

John, what are we talking about here? I specifically referred to the ability of analog consoles to route tracks to multiple buses at once. You can do this in sonar using sends. But generally you have the ability to do this by assigning the track output to multiple buses, not just one. The whole analog comparison is kind of silly I know as Sonar is a lot different in many regards. But at the same time it really seams that's what the Sonar console is modeled after. And it would be nice to have this ability. Is it needed? No! Would it be nice? I think so. 

like most I too go tracks to instrument busses to master
I also have my click sent to a buss and direct to output (missing the master buss) allowing for better levels and control during playback/recording.
Sends run to FX busses which fed to the master
will also run a send or two off the master to busses with compressors running the output, when I need to do some parrallel compression

Risking the possibility of becoming a thread killer once again, I offer you a different approach.
Below are quotes from my own posts in another thread, and there's definitely a merit in routing tracks directly to the master bus.
Note: This is a busy man's approach (or rather a songwriter's approach) and thus not recommended for everyone.  

Because, for example, if you want to apply the same delay (same plugin, same delay time, same feedback amount, etc.) to both the lead vocal and the lead guitar (but not to background vocals and rhythm guitars), you'll have to create two effect buses and insert exactly the same delay plug with the same settings to each of the buses, and then route one of them to the vocal bus and the other to the guitar bus, because if you route them to the master bus, the balance between the dry sound and the wet sound won't be the same when you adjust the volume faders of the lead vocal/guitar tracks. 

if you just think it would be convenient to adjust the volumes of multiple instruments with a single fader, I suggest a better way, which is simply grouping all volume faders of tracks of the same instruments (group all volume faders of all guitar tracks, for example). 

To give you a better example, you might want to use only one instance each of three reverbs (e.g. room, plate and spring reverbs), two delays (short and long delays) and one chorus plugin to make the setup simple, and want to apply them to different tracks (room reverb to distortion guitar, organ and bass, plate reverb to lead guitar and some synths, and spring reverb to lead vocal and clean guitar, long delay to lead vocal and lead guitar, chorus to clean guitar and electric piano, etc.), this volume fader grouping approach can achieve it quite easily (and you can adjust the tracks' volume levels separately by holding Ctrl while tweaking the desired fader, which is why I said "why not take the advantage?"). 

I consider this approach more flexible, and the only bus I set up now is the drum bus (to apply saturation, mild, total compression, etc.). But what approach to use really depends on the user and the project, so just take this approach as one option that could be of help. 

http://forum.cakewalk.com/tm.aspx?m=2831046

If you're wondering about collision between volume fader grouping and each volume automation, you can instead use clip automation or the Gain parameter for volume automation, or group the gain parameters of tracks you want to group for total volume control.

Sonar offers many features and options, hence many ideas.