Well I use Voxengo SPAN (which doesn't have a Linux version) with the mastering preset.
The reason for using the mastering preset is that it sets the response to be quite slow and averaged out with much less detail in it so you won't see harmonics spiking up everywhere. This is exactly what you want when using it on the master bus because it makes it a lot easier to see the big picture which is what mastering is all about. We're not concerned with momentary peaks in frequencies here unless they are very high peaks.
The other thing with this preset is that it sets the slope to 4.5db per octave. This is really important because it will significantly alter the presentation of the frequencies. A 0db slope would represent the frequencies at the actual level they're at but human hearing is more sensitive to some frequencies than others and so setting it to 4.5db tilts the slope and more closely resembles how we actually hear them (Play with this setting and you'll see). Not all spectrum analysers will use the same slope by default and so you can see very different results depending on which one you use which can leave you feeling very confused. A slope of 4.5db means that when a mix is well balanced you will see roughly the same level across the entire spectrum; All frequencies will be moving around the same volume level as eachother so it all looks quite flat which makes it very easy to understand 
This is not a solid rule though as it totally depends on what's going on. For example, maybe your music has no bass instrument and is just an acoustic guitar with a vocal. Well in this scenario it'd be totally normal to see less activity happening in the low end; It'd be more mid-range focused which is fine. So just keep this in mind! A fully fleshed out song with all sorts of instruments playing in all frequency ranges at the same time will, however, likely look more flat across the whole spectrum when it is well mixed (although looking flat on the analyser does not guarantee the mix is good), with the frequencies rolling off at the very top and very low end of the spectrum; Somewhere around 20Hz & 20KHz, for example, as there's no point having frequencies outside of this range because no one can hear them..
So to summarize:
- Select the mastering preset on your spectrum analyser if it has one. If not then you will have to read the documentation or seek advice from someone else who uses it to be able to set it up for this task.
- Set the slope to 4.5db.
- Mix by ear but also use the spectrum analyser as a second reference and try to get the mix looking fairly flat across the spectrum, assuming that's what is appropriate for your mix. But as I said; sometimes this is not the right thing to do.
- Now check your mix on different devices in different locations and hopefully you'll see it has translated better!
As for sub bass; Yeh I use that trick all the time. I like to put a subtle sub underneath the main bassline just to fill out the low end. It's one of those things where its presence is felt rather than heard. I also like to compress the two together on a bus as I find that helps glue them better. You don't always need a sub though. You'll start to realise when you do or don't need it after some practice
It took me a while to write that so I hope it clears things up a bit..?
