ULTRABRA
It sounds much easier simply to increase the Klanghelm Volume Control knob by a few db (or whatever it takes).

Conceptually, I also don't think it makes sense to use the meters to adjust level.  If you do have to insert a gain stage, it would be clearer to use something like the free Blue Cat's Gain plugin from Blue Cat Audio.  To streamline your gain staging it might be a good idea to go back to the very beginning and adjust your levels consistently moving forward, grouping those tracks that you want to maintain in relation (by linking or through a new buss) as you go. 

Love Blue Cat's gain plugins! Especially for volume-matching vocal clips.

I've been playing around more with these VU meters.  What strikes me most is how "approximate" it seems to be - the needle bouncing around so much, how is it possible to have any degree of accuracy?  Is it not better to use something that gives an EXACT RMS reading for the track?  For example, I inserted the Sonalksis FreeG, and it gives me an exact RMS value for what I'm playing back.   The VU Meter needle varied wildly, and the results didn't match.   Maybe I'm over-thinking this ??

When a VU meter is swinging wildly it can mean the ballistics of the meter are telling you have not tamed the dynamics perhaps. Once you start controlling dynamics nicely the VU does not swing anywhere near as wildly. It changes quite a lot even with mild compression.
 
What VU meters are you using as a matter of interest. I must admit though that the real thing and real nice expensive ones at that still move in a nicer way and in a more fluid manner. I think when it comes to ballistics the actual real meter is probably very hard to beat. (I suspect it may be harder to emulate than what people think)
 
I am spoiled in that regard because I use the hardware meters most of the time and I know it is not something many are able to do.
 
The Klanghem meters are pretty good though and close. But they can still tell you what is going on. When the meter is moving nice and just hitting 0 dB VU then you know the rms level is at the reference level. So they can be accurate. The problem with other types of VU meter and even ones that show you accurate rms readings they fall down in the ballistics department and don't let on as much as the real ones do re dynamics being a little out of control.
 
Read the VU response of a real nice mixed and mastered CD. See the difference. You will have a different opinion about them then. (preferably not a smashed master either, one that has some nice dynamics in it) Adjust them so they are just hitting 0 dB VU most of the time and observe. Steely Dan's 'Everything Must Go' looks great on a (good) VU. A great example.

ULTRABRA
What strikes me most is how "approximate" it seems to be - the needle bouncing around so much, how is it possible to have any degree of accuracy?  Is it not better to use something that gives an EXACT RMS reading for the track?

If you're keeping things down in the -20db to -14db range, it really doesn't matter what the exact value is.
I like Jeff's idea of getting into the dance of the needles as a way of getting information on the dynamics of the music.
I'm sure there are other ways this could be done.  I find VU meters easy and relaxing on the eyes and brain.  Hypnotizing really.

 
The Klanghelm meters have to be adjusted to the reference you choose.  If you choose -20db, and set you meters accordingly [not with the gain knob in the middle, but the (red) "VU" value at the bottom right of the meter, click and drag up/down], when the needles just touch zero the level is really at -20db.  So the zero on the meter is your reference.
 

Ok, for me the effect buss's should always be around -18db RMS or VU meter.  If you are not going back out of the box when mixing...don't worry about overs.  That will be controversial to some, but trust me...don't worry about overs, unless you plan to use a lot of outboard effects.
 
Now when it comes to sub mixes, first get the mix in order.  So work out what tracks need to stand out, and what ones are in the middle, then the rest.  Get the main tracks, nice and loud...so near 0 VU, the middle ones -6ish, then the rest.
 
Now, and again this will be controversial, just let the instruments you want to sub-mix hit the channel.  Don't worry about overs, unless again you're going to feed the sub-mix into an outboard device.  It doesn't matter how far over you go...of course they're still limits...and you will hear them.  Use the trim knob, and if you like, the tape sim...which is what I do, and line the sub-mix up to around -6 VU.  Then mix the buss, use gain staging between effects by all means...but don't worry if you're going over, unless you're going out of the box.  Then just turn the sub channel down to fix the mix.
 
This is how I mix, my websites are all in my profile, if you don't believe what I am saying, or someone here says this is wrong.  I stand by what I've written.  This is the digital audio paradigm, and goes against what most people would tell you...because for the most part they have come up through the analogue paradigm.  Not that any of the contributors here are wrong per se...it's just mixing in the box requires a new way of thinking.
 
Peace Ben 

Jeff Evans
When a VU meter is swinging wildly it can mean the ballistics of the meter are telling you have not tamed the dynamics perhaps. Once you start controlling dynamics nicely the VU does not swing anywhere near as wildly. It changes quite a lot even with mild compression.
 
What VU meters are you using as a matter of interest. 
 
Read the VU response of a real nice mixed and mastered CD. See the difference.

Hi Jeff,
 
I understand what you mean by the dynamics may need compressing --- but as I put the VU meter as the first plugin in the chain, its coming before any compression.  Isn't this how it should be done?
 
I'm using the Klanghelm software meters at -18.
I can understand how a final CD mix will have a more smooth response, but surely that is to be expected compared to individual tracks which will have relatively more dynamics?
 
I'm still trying to figure out how best they can work for me.  I've had good examples where I can see the VU meters going +4 in the red eg on a bass track - here I can easily lower the output on the soft synth to get it around 0VU and then raise the fader at output to get the right balance in the mix.  But, on the othe rhad, I'm having a problem with for example sustained string patches which only get the VU flickering at -10/-7 even when the output is maxed, even though the volume output seems good.   Similarly, fast sequences eg repeated 16th note patterns sometimes also show -10ish, even when the output volume is good and peaks even -2dbfs.    If I raise the output there to 0VU I'm going over 0dbfs with peaks ...  I guess that could be because of the fast transients, but that does not explain the sting patches showing very low in the VU.   
 
 

You have brought up some excellent points.  Monitoring incoming raw signals or the outputs from VST's directly with VU's are going to give you the wildest variations in VU meters for sure.
 
If a bass VST is averaging +4 db higher than 0 then yes what you are doing is the right thing.  Lower it so the output is hitting 0 dB VU.  I aim to get all tracks at a similar ref rms level.  Their final output is adjusted on your channel fader later in the mix to give you the right balance in the mix later on.
 
Good example from your sustained string patches.  If the level of a VST is consistently hitting -7 with its output maxed, what I do is insert gain after the VST and add 7 dB so I am getting my 0 dB VU again.
Track ends up at the ref level as it should.  In the mix however the track is once again set for the right balance and its channel fader may be down at -15 dB or so according to your ears.
 
This is good because you have shown here how wildly variable the outputs from virtual instruments are.
Some of mine eg Korg Wavestation can be so low I have to add 10 dB of gain to some patches just to get them up to 0 dB VU.  Others I have smash the VU hard over and I have to turn their output level way down just to get the meter to sit nice at 0 dB VU.
 
If you are wondering why I add gain to bring some VST's up to zero it is because once you have done it you won’t have to deal with it again in the mix.  You will always have a healthy amount of signal to play with in your mix later on.  Not so if you leave and record it way down low.  You will have to add gain at some point later so why not get that out of the way.
 
Now another excellent example with very short staccato string lines.  Some patches will barely make the VU move yet be close to 0 dB FS.  I have always said you need your peak metering too and in cases like this your peak meters rule.  You can't let the strings clip so in these cases I just set them so the max peak is good safe distance from 0 dB FS eg 6 db or so.  Dont sweat the VU's in this case.  Very snappy percussion sounds are in this category too.
 
What I do though in cases like the short string sounds is later I open them up in an editor and limit the tops of those peaks down maybe 6 or 7 dB or so then I add 6 dB or so to the whole track. That brings the rms level up without changing the staccato nature of the sound.  You will barely hear the difference except you will now have a decent rms level going on as well as the attack transients.  I prefer to do it this way.  You could always put a limiter after the VST and add gain but then you might screw up the sound.  At least with the editor you have complete control and you can always go back to the original recording if you want.  And as you say even though the rms level of that sound might be down compared to others if you end up hearing it OK then it does not matter really.  As long as you have enough to get that part heard in your mix.
 
With lots of drums I have found by the time you send a whole bunch of very transient sounds that are not making a VU move much individually to a drum buss, there will be enough total rms level there to move a VU again back up to the ref level.
 
VU's are great but you also need your peak metering and you use both at the same time.  A lot of stuff though can be handled with the VU alone and the peaks are only a short level higher than rms.  But you also have to be able to accept those signals with very low rms levels and high peak levels.
 

Thanks again Jeff ... I had just started to experiment with adding gain on those string patches - now what you said makes perfect sense there.
 
When you say "VU's are great but you also need your peak metering and you use both at the same time" :  I also just downloaded the demo of PSP Triple Meter, which shows VU, RMS, and Peak meter all in one.   I found that the RMS meter there, also calibrated -18, is much easier to see - those fast sting spiccatos are showing 0 here, when the VU meter is still -3/-5 down.   On other patches the VU and RMS meters are more or less same.   I'm not actually sure the difference - I thought the VU meters were measuring the RMS .. anyways, that PSP RMS needle just seems to be more easy to read, less fluctuation so easy to set for 0.  
 
I'm going to use both for a while, I guess most important is not to let it get over 0VU too much, use gain where necessary, and for fast transient patches rely more on keeping the dbfs peaks in check at not past -6.   I'm trusting that will get me most of the way there, safely ... :-)