With digital files clipping at 0dB there is an inherent limitation that we have to work against. There is only so much musical information we can fit into a file and that cutoff point is the 0dB ceiling. Once a track is exported to a WAV or AIFF file, any information that is higher than 0dB will be chopped off, leading to a distorted, 'in the red' clipped sound. Usually undesirable.
Limiters, are the default tool to restrict the signal below 0dB, keeping it in a safe level, but too much limiting can reduce the impact and punch of the dynamics (at best) or make a track sound spongey and lifeless (at worst). Limiters are usually always applied to tracks as the final protection against clipping in the mastering stage, so it's worth having a bit of understanding how they will effect your mix if you want to sound loud and proud.
Limiters are essentially really fast compressors that prevent the loud peaks of a sound from going over the 0dB threshold, with minimal distortion (at least that's the advertised function, limiters come in many different designs of varying quality; some distort more than others).
So when your aim is to have a loud sounding track that still sounds good, the real question to ask is:
How do I create a mix or master that will interact with a limiter in the best possible way?
When we reframe the loudness question like this the answer becomes easier. Here I suggest the most important things to look out for as your crafting your mix.
While I definitely don't advocate worrying about final loudness throughout the mixing process, this can be a distraction and a barrier to making what should be artistic decisions at this point. But if a certain degree of competitive loudness is your goal then it's a good idea to devote a small bit of time during mixing to check for some of the ideas mentioned here and to check through a high quality limiter just to see how things are going to change.
The first thing is tonal balance. This is the most important and the most common mistake I hear as a mastering engineer when it comes to loudness barriers. If you only take one thing away from this article let this be it.
As Fletcher & Munson have taught us, the ear is most sensitive to upper-mids/lower treble frequencies. Bass frequencies, especially subbass (because the ear can't hear them as easily as the lower treble) usually need to be higher in the mix to be heard. Also bass frequencies have longer wavelengths so they tend to eat up headroom and push other sounds into clipping and distortion very easily. (More on simultaneous sounds later.)
Once you apply a limiter to a very bassy track the bass hits the limiter first and starts that characteristic ducking sound, pulling away all the other musical information on each bassy section. When this happens it becomes impossible to get the track up to a competitive loudness level. So mixing your bassy sounds too high can be a major barrier to loudness and clarity in the final master.
Then again, you might want a bassy track. You don't want to be 'ruled' by loudness, the music should come first. Always. So if you want a big bottom end on your song then how much bass is too much? It's actually easy to know how much.
If you make the tonal balance of your mix roughly the same in terms of bass/treble as a few references, you're doing plenty to combat against the problem of too much bass. Remember it's very important to do this by ear and not by visual scopes. The mastering engineer can further tweak this balance but if it's way off where it needs to be, there's only so much a mix can be EQ'd before it starts sounding very strange. As usual it's better to get this right in the mix so aim for a similar tonal balance to a few selected references.
And of course, the treble/upper-mids area has to sound good. In loud masters this region is usually very present and on show so effort needs to be put in to make this sound just as you want it to. We've covered this in our article, 'Strategies to stop treble from hurting' so we won't go into detail here but to re-cap, if you can make this frequency region sound good when the volume is turned up and incorporate the other techniques covered here, you'll be well on your way to a loud track that still sounds good, and will have less unpleasant surprises when sending your tracks off for mastering.
Highly dynamic sounds are by their very nature hard to squeeze into a loud track as the tall peaks of transients usually present in a dynamic sound literally don't fit in below the digital ceiling of 0dB. What usually happens is some kind of compromise compression or clipping that will reduce the peaks of the dynamic sound such that they stay below 0dB. There's no getting around the 0dB limitation, so if you're really keen on highly dynamic sounds for your track then you might have to become okay with a final master that is less loud than some of the louder tracks that are out there (and that's okay).
This can often be a barrier to a loud sounding mix or master because it's easy to overdo. If you like this sound then make sure to reference against some tracks that have this kind of drum sound and strike a good balance between snappy attack and the sustain portion of the drums. Don't overdo it as you'll find that when it comes to mastering stage the overly large transient attacks won't come through as you had planned if you're going for competitive loudness.
The human voice is hugely dynamic. Loud portions are great for dynamic tension and the reminder that you are indeed listening to a human! (And consequently a reminder of what it means to be human, but that might be getting a little philosophical.)
However, have a listen to the vocals on mixes that you love especially with dynamic performances. What you might be surprised to hear on listening closely is that a lot of these are mixed to have quite an even volume, with the intensity of the vocal performances carrying the loud portions. When done well, the ear still hears this as loud (and the hairs still stand on the back of the neck as desired). Keep the vocal dynamics controlled and it'll be much easier to mix the instruments accordingly too.
It might be okay to stack a hihat on top of a kick drum since even though a kick takes up a lot of space in a mix, a normal hihat doesn't need much room at all. But stack a thick snare hit on top of a kick drum when you're writing a song, by mastering time you might find that this combination might be too much for a limiter to handle, especially if there's a few more instruments playing there too. Loud songs, at least ones that still sound good and loud, tend to be arranged quite cleverly so you might have to go further back than just the mix if you find your song has a problem like this.
Again, this is probably something to keep in mind (but in the far, deep recesses of the back of the mind; it's lower priority). You want to be making good music first and foremost, not constructing some sound experiment in loudness capacity for academic purposes. Write the music you want first, later if there's an issue here with simultaneous sound combinations pushing things over it will easily show itself if you apply a limiter or a clipper to the track to test how it will react.
Why do some of recent 'EDM' tracks sound so loud yet don't distort?
The answer lies in the choice of sounds. These kinds of songs are often a swathe of white noise ducking and pumping mixed in with other sounds that are also broadband in nature. If white noise is the most broadband sound then the most resonant sound would be a single sine tone. Let's go forward with those two distinctions. Broadband sounds vs. resonant sounds.
Noise, especially white noise, is loud. Not necessarily loud in a good way, white noise hurts (it has a good deal of all frequencies by definition, including the upper mids/lower treble). But if you imagine all sounds being made up of a noise component and a resonance component, the ratio of noise to resonance will be different for any given sound. For example the sound of a waterfall has a large noise component and very little resonance (you can think of resonance as 'musical note'). A piano has a small bit of noise and is mostly resonance (in fact solo piano is possibly the limiters worst enemy out of all the common instruments). A sax, played quite breathy would have quite a large noise component and less resonance than a piano. With some synthesised sounds, you might have complete control over the amount of noise present in the sound, which plays into the hands of the EDM producers who want the loudest track known to man (sorry to disappoint guys but Iggy's 'Raw Power' is probably still louder).
The EDM world is probably an extreme example, but the thinking still holds. if you have resonant sounds in your track, piano, hi-Q filter sweeps, sine-like tones, try and make sure they are dynamically controlled and mixed well in the track, sitting firmly in the bed of instruments. This way they won't hit the limiter too hard and cause it to distort or duck the rest of the sounds.