Carnage
Don't sweat it. It takes a long time to learn to mix and master.
I think you've pretty much narrowed it down to two issues: final product loudness and the much fuzzier issue of the art of mixing and balancing instruments.
Let's continue working backwards and take the final mix loudness first (since it's easiest).
It's usually easy to do a simple
normalization of a mix. At its simplest, normalizing means making sure that the loudest peaks in your mix fill up the allotted dynamic range... that is, since 0 dB is the loudest digital signal possible (and lesser volumes are measured as negative numbers), you want your peaks to be as close to 0 dB as possible without "crashing" (what happens when you feed to great a signal into a digital converter... in essence the wave just 'flattens' against that 0 dB ceiling and becomes one big (or a series of smaller) square wave... producing the dreaded and evil Digital Distortion. So you want your peaks to just touch 0 dB or come just under it. And that's easy to do, straight multiplicative math. If your unmastered mix peaks at -5.5 your software's normalization function will simply add 5.5 (or maybe 5.49) dB to the sound across the board. Your mix will sound precisely the same only louder.
Unfortunately, that's often not enough -- because the average volume of the mix is too low. You can't normalize it any more because you're peaks are already topping out. So, then, what you have to do is do some compression to lower those and some gain compensation to bring everything up so that the new peaks are just barely maxxed out.
Sometimes, that's still not loud enough to compete with commercial mp3s or CDs. (And maybe you shouldn't -- many commercial releases seem over-compressed to many people.
Squashed is the term of art. You usually hear the problems in the cymbals first, they lose their sparkle and sheen in the mix... as things get worse they can almost disappear.)
Rather than squashing the stuffings out of your mix to get that relative competitive volume, you might consider a little careful thinning at the bottom of the frequency range. Because of the physics of the thing, low frequencies suck up the most power and volume. But you may have subsonic information that most listeners will never hear -- or that listeners might not even want to hear [distant 787 jets, trains, trucks, subways]. Many mixers carefully roll out varying degrees of sub bass to give more room for the rest of their mix in the finite dynamic range they have to work with. It's safe to say that below 30 Hz there is little of value or interest and that almost no consumers have gear that can reproduce these frequencies anywhere near accurately. The range between 30 and 50 Hz is more of a judgement call. Most rock doesn't have a lot going on down there, but orchestral music, and dance/synth stuff often does... you may have to balance your needs when making these decisions.
When making these decisions you have one very important tool -- your ears. You may also have spectral analyzer software (typically the little moving light bar freq display you see killing time in Mp3 players. The 'graphics mode' of the Timeworks EQ that comes with Sonar 2 deluxe (or whatever they call it) has a realtime spec analyzer that really helps.) But the final arbiter is your ears. (Which is why a good monitoring situation is critical. If you don't have monitors you can rely on, you can still take your mix to as many listening environments as possible and also use a CD (or CDs) you think sounds good and you're familiar with as a 'reference' to try to match to.
It seems like we've already sort of meandered back to mixing...
As I'm sure you suspect, no one can teach you mixing in a few paragraphs or a few lessons. A lot of it is just doing it over and over again. You practiced your guitar or keyboards or singing to get good and I'm afraid you'll need to practice mixing, too.
But there are a few things you can do to make it easier. I mentioned the reference mix thing above and that can be very handy -- especially if you're working in an unfamiliar monitoring environment.
Another thing that may help is to think 'visually.' It's easy to do with the left-right stereo soundstage... but you can also train yourself to hear "chart" differences in frequency / tone on an imaginary 'vertical' axis. You should further get in the habit of hearing your mix in terms of front-to-back depth, as well.
You've probably read that it's important to make room for each of your elements in your mix space. That means not stacking similarly timbred instruments in the same place. (With the possible exception of the bass and kick drum, which a lot of mixers like to think of as a team.) But don't put a mid-rangey electric piano in the same space in the mix as a mid-rangey organ... or don't put an upper mid-range guitar in the same place as a vocal. They will tend to mask each other and distract. But if you give them each their own place in the mix (usually left to right but also front to back to a certain extent) they ears have a much easier time separating them and making sense of things.
With regard to when to fold your MIDI and soundfont instruments in... good question. You probably have a bunch of guitar and vox on audio tracks and a bunch of keyboard, bass, and drums on MIDI. At some point you either mix it all up at once into one big audio mix -- or you bring the MIDI tracks in as one or more pair of audio tracks and then use Sonar's clip mixing / audio export to mix it all together as a stereo mix.
If you're using a number of voices coming out of a wavetable synth (and your Sound Blaster sound font synth is just such a beast) you may get better luck by recording individual MIDI tracks or small groups one at a time, applying EQ and FX to them as appropriate. It's easy to mix all your MIDI tracks into a stereo pair (or fold them into the mix) all at once -- but I've found that you often get this flat 'music-in-a-box' feel when you do so. Part of the reason for that is that you are of necessity EQing and FXing all the simultaneous parts the same way. If you record each track separately you can massage and nuance it separately... maybe adding a little extra reverb or echo or flange or whatever to certain parts in a way that wouldn't be possible if the piano, organ, drums, bass, etc were all piling out the synths stereo output at once. (OTOH that will put an extra burden on your system as you'll end up with more audio tracks.)
For years, because of the peculiarities of my setup, I did usually record a single mix of my MIDI and audio tracks (I had 8 audio tracks in and out and I have a bunch of separate synths, drums, even a separate bass module since I started a long time ago when there was no such thing as a soft synth -- or even SoundFonts.) But the last mix I did I ended up doing precisely what you wondered about -- I recorded my MIDI tracks as separate audio tracks (I also use a lot of SoundFont drums these days... this gave me a chance to add a little subtle compression to the drums independently to smooth things out).
I routed everything through a Sonar virtual buss and inserted a stereo EQ and compressor (in that order. I used the Timeworks EQ and compressor that came with Sonar XL [that's the name!] but one could just as easily use any other DX plug-in. I brightened the top end around 10K just a tiny bit [I'd already brightened the drums separately some] and I set the Timeworks FX to engage the 'brick wall' limiter just under the 0 dB mark and then tinkered the other params until it still sounded like it had some 'air' in it -- but the meters were tightly bouncing in the top 12 dB range [actually it was in the top 6 dB quite a bit])
I'd previously been using 'offline' mastering tools in Sound Forge (they have a great 'smart' normalizer that can be set to automate final compression pretty well). But in a non-real-time environment like that you find yourself working through trial and error and maybe settling before you should. Setting the final compression and eq 'live' in Sonar worked great for me.
(I actually ended up recording my SF drums onto separate MONO tracks instead of a single stereo track so that I could reign in some of the excess stereo spread built into my chosen drum set. Instead of setting the individual pans at "7 o'clock" and "5 o'clock" I set them around 10:30 and 2:30 and it sounded much more in keeping with the small chamber rock thing I was doing. Even when I'm working with electronic drums I always try to visualize the drummer and other instrumentalists as I'm mixing.)
Hope that helps! You might also check out the links I've collected on the Audio Resources page I have linked below my signature.
