Our loudness perception is relative. Some sounds we think sound loud, actually don’t sound loud and visa versa.
Our ears are most sensitive in the midrange, so when lows or highs are peaking we don’t perceive this as loud, but when something in the midrange is peaking we do perceive it as loud. Therefore when we apply saturation to a signal, adding overtones in the midrange mostly, we tend to perceive that sound as louder. Overdriving 808 kickdrums for example is a populair technique these days. It makes us believe that kickdrum sounds louder, but only additional harmonics are added. The peak is not louder. It’s a trick that is also useful because it “translates” the clean sine wave 808 kickdrum sound to devices which are not able to reproduce the super lows. Overdriving a 808 kickdrum makes it more midrangy sounding without loosing any of its booooooom. Even on a smartphone you will be able to hear the kickdrum because of the added midrange.
And the sound is actually indeed louder when we add transients to a signal using saturation. Those additional frequencies make the overall RMS/LUFS values go up. The peak frequencies will stay the same when overdriving a signal (added overtones are not peaks). So the RMS/LUFS is going up and the sound will be louder. Saturation causes the signal to compress (making RMS/LUFS values go up) but without using a compressor. This is the magic of saturation. Adding overtones, adding colors to a signal. Playing Vincent Van Gogh for sound. LOVE it!
But be warned! When RMS/LUFS goes up this means the sound will be less dynamic. So using saturation will always be a compromise between added loudness and loosing some of the dynamics. Music needs dynamics. And it needs compression. When you overdo it the sound will be too much distorted, too much compressed, or simply put: boring sounding.
For many years there was this thing called The Loudness War. Most mastering engineers where pushing the tracks into a brickwall (read: hard clipping using limiters, creating ODD harmonics). Some still do. That sounded awful because of the lack of dynamics and the hard clipping sound they created using brickwall limiting. Hard clipping can be useful though, on bassdrums for example. The 808 for example with its sine wave tone doesn’t have much overtones. When we overdrive this with hard clipping the sine tone will turn into a square wave tone. So we’re tweaking the 808 oscillator in a funky way by overdriving it 🙂
During the loudness war rule number one was: there are no rules. So everyone was trying to compete, to sound louder than someone else’s track (read: getting the loudest RMS, clipping the master like crazy!) using brickwall limiters.
I f#cking hated that sound!
Follow the rules!
Late 2010 EBU published the first version of its Loudness Recommendation EBU R128. Soon many European broadcasters started using that norm. Basically EBU R 128 recommends to normalize audio at -23 LUFS (from beginning to end). LUFS is like RMS. This means that the average loudness of a track should be at -23 LUFS when you mean to broadcast it on radio.
What about streaming?
iTunes Music, YouTube and Spotify don’t follow the same EBU R 128 dynamic rules. In fact they have created their own loudness rules:
|iTunes Music||– 16 LUFS|
– 13 LUFS
– 14 LUFS
This means that when you mix your track based on the iTunes norm at -16 LUFS, this norm offers you a potential more dynamic range than Spotify. A track that is mixed based on the Spotify -14 LUFS norm will be lowered in volume when played in iTunes. What rules should we follow? That’s a tough question to answer. Popmusic and radio conversations need compression, so the question these days still is: how much?
Some people simply say: don’t worry about it, mix it as dynamic as you want. But this might mean that a super dynamic at -23 LUFS will sounds too soft because to avoid clipping the tracks will only be attenuated but never get positive gain. The other way around: a loudly mastered mix at for example – 8 RMS/LUFS will be attenuated in volume and since the average of an album will be used, maybe some tracks will end up to sound even softer than other albums.
Keep in mind that most streaming upload services like Distrokid will ask you for just one file to be uploaded to all the services. So creating separated mixes for like iTunes and Spotify is not at all convenient.
A clear answer is hard to give, but I guess the people who say “don’t worry about it, make the best sounding mix you can” are right. Setting the final LUFS level is just a matter of setting the gain for the whole track differently.
What to choose?
EBU R 128 is a fantastic norm for radio which has a wide dynamic range to offer. But do you need such a lot dynamics in your track? Probably not.
The iTunes norm in my opinion makes a little more sense -16 LUFS. It is in my opinion the best compromise for modern productions.
P.S. what loudness norm is SoundCloud using? Deezer? Tidal? Google Play? I would love to find out!
The Mastering Show podcast recently did an interesting episode on loudness. How to make something sound loud?
UPDATE: Some small errors were fixed. As for a norm for mixing maybe -14 LUFS is currently the best target level. See also my newer article on this: ‘Dutchman Eelco Grimm has examined 4.2 million albums (!) on loudness’
Also published on Medium.