Getting a grip of the signal flow:
-
preamps (interface or external) have a sweet spot, probably met around -18 dbFS in Cubase (see documentation of preamps/interface)
-
recording to 24 bit or 32 bitFP (32 just makes sense when recording with plugins or extended usage of offline processing but also doesn’t hurt much, see other thread) offers lots of headroom indeed
-
the internal headroom of the Cubase mixer is around +1500 db (of course your converters’ headroom isn’t) so internally nothing can clip, but…
-
…plugins have a sweet spot regarding leveling just as the preamps and converters - this is different from plugin to plugin. Serious emulations of vintage equipment certainly have a standard they’re calibrated to (see documentation) - good idea to make wise decisions of input and output levels here (= not abusing plugins, unless it’s a consciously made artistic decision)
-
due to the internal 32 bitFP processing the Cubase’ channels can exceed 0 dbFS without damage to the signal but as there are a lot of them summed it makes sense to keep everything far below just like on an analog desk (ok, analog sounds ‘hotter’ when driven hotter, digital doesn’t without extra efforts like saturation tools/console emus etc.)
-
of course the master out(s) should never exceed 0 dbFS as the master level hits the hardwares’ converters
That’s all the magic: keeping things in the sweet spots of whatever hard- or software (plugins) used. Otherwise the master out becomes sort of a bottleneck, with everything hitting against the brickwall.
I go around -12 to -10 dbFS on the way in. This has proofed to be a well working leveling on my hardware. It happens that some signals might be hotter but also lower. As long as the pure file sounds good, everything is good that far. For the mix, the channel gain can be lowered or raised to get to whatever your idea of an ideal level is (-18 to -10 db seems like a good range…).
When mixing I’ll do some gain staging anyway, almost always using Slate’s VCC usually as the first plugin - I set the gain of every channel to hit 0 dbVU on VCC’s meter. If I don’t use it, it depends on the plugins I use. Most ‘formal’ EQs are pretty forgiving (probably working in 32 bitFP internally), ‘vintage’ EQs might produce different colorations when driven more or less hot. Dynamics need proper gain staging (input levels) anyway - technically. ‘Non-proper gain staging’ can be the method of choice as well if the result is what you’re after.
About busses: same story, usually below 0 dbFS, also depending on the potential bus plugins. If the channels feeding the bus have a good relation but are too hot/too low for whatever reason - that’s what the group channels’ input gain is for. Same thing for the master, which is just another bus.
Oh - I was late to game, iBM already said some wise words 