Note: there is an update to this article (August 2018).
In a previous post, I covered how the Loudness War started, and why it's bad for the sound of music from increased distortion and decreased dynamic range. Fortunately, there's a glimmer of hope in that many of the popular streaming services are turning down songs that are too loud in order to even out playback volume for listeners. As of this writing, YouTube, Spotify, Apple Music, and Tidal all turn down songs that are too loud, eliminating any benefit from exceedingly loud mastering.
That's great, but if you don't know how loud to make your music under the new rules of the streaming services, it won't do you any good. That's what we're covering today.
I'll warn you right now: this probably won't be fun to read. I'll have to write a lot more digits and acronyms than I'd like. But the volume of your song can't be taken back after release. And getting your head around this one sticky topic will permanently help you increase your music's volume if you master too quietly, or more likely, permanently improve the sound quality of your music with no penalty for the vast majority of listeners who listen through streaming services, radio, and modern media players.
Identifying Your Target Loudness
This is where the supremely helpful Ian Shepherd (British mastering engineer and author of the Production Advice blog) comes in. He has a very nifty chart outlining the loudness recommendations from the Audio Engineering Society, along with the target loudness of four streaming services, and how each one handles songs that are louder and softer than the target loudness.
Coming to Terms with Loudness
That chart is really helpful if you know what LUFS are and how to measure them. Otherwise, it's just further muddying the waters!
As you probably know, the standard unit of loudness is the decibel, or dB for short. The decibel was established as "the smallest unit of loudness that a person could hear", though most engineers can hear a difference of just a tenth of a dB. A quiet fan in your computer might whir at 35 dB, a loud restaurant may be 80 dB, and a loud concert may be 110 dB. It's not a linear scale: 76 dB is twice as loud as 70 dB, and 82 dB is four times as loud as 70 dB. An amplifier requires double the wattage to play 3 dB louder; an amplifier pulling 25 watts would need to pull 100 watts to play 6 dB louder. The loudest possible sound by this scale is 194 dB, limited not by the sound source, but by the physics of air and Earth's atmospheric pressure: any louder, and the sound would behave closer to a shockwave than a sound wave. That said, sounds much louder than 194 dB are scientifically possible underwater or in the atmospheres of other planets.
In the digital realm, it's much easier to achieve the loudest possible sound: decibels are relative to "full scale", or the loudest that a digital file can store a sound. In a session, 0 dBFS (0 decibels full scale) is that max ceiling, since peaks can't exceed 0 dBFS without being written to the file as pure distortion. As an example, recording vocals at quiet levels might have you record in at -20 dBFS, and you may turn them up by 5 dB in your session to play back at -15 dBFS. Everything we do in our DAWs is in the negative scale, working down from 0 dBFS. It's just how digital audio works.
Loudness in music is tricky to measure. Sure, we can measure the peaks according to dBFS, but that doesn't really tell us anything about the average loudness of a song, but rather just how loud a single moment was. Classic VU (volume unit) meters measure average volume levels over time, but they often don't respond the same to different styles of music or even different songs of the same style of music. And how much time is averaged can vary the reading significantly. The same problems apply to RMS meters and their readings.
Enter the solution to this problem: LUFS, short for Loudness Units relative to Full Scale. The history of LUFS can be traced back to the International Telecommunications Union from as early as 2006, and more recent revisions have clarified and expanded upon what has gradually become the new unequivocal standard for average loudness measurement.
Fortunately, it's easy to understand: one LUFS unit is the same size as one decibel, so understanding the scale becomes simple if you're familiar with any traditional meters relying on the decibel for scale. The value of the scale, however, is two-fold: there's a lot of math happening behind the scenes that makes LUFS a much more accurate and consistent measure of average loudness than VU meters and RMS meters, and it's been widely adopted by streaming services and government broadcast regulations alike as the one metering algorithm to pay attention to.
Also, fortunately, reading it is easy once you get your head around it. LUFS can be tracked by the momentary value (averaged over 400 ms), and by the short-term value (averaged over 3 seconds). But the integrated loudness is tracked and averaged over the entire length of your program. Meaning that if you are working on a three-minute song or a forty-five minute TV episode, the integrated LUFS is a single value that gives you the sum of loudness for your entire project.
A lot of different brands of plugins are now offering tools that can measure LUFS. iZotope, Waves, TC Electronic, MeterPlugs, Dolby, Avid, NuGen, Mastering the Mix, Klangfreund, and HOFA are just a few. Youlean even makes a free plugin that measures LUFS, though I haven't used it personally. I use iZotope Insight, which I share more about here. Pick a plugin to do the job, and find where the short-term LUFS value and the integrated LUFS value are displayed. These are the primary numbers to pay attention to.
The Advantages of Each
Since LUFS-integrated captures the loudness of the entire song with one value, it's easy to read. And since streaming services use algorithms closest to LUFS-integrated to determine how much to turn down loud songs, there's a built in argument for exclusively focusing on integrated LUFS.
When using LUFS-integrated, targeting a value of -13 to -14 for your entire song makes finalizing your song's loudness easy. At these levels, you're likely not compressing so much to have audible issues, but you're also not ignoring useful loudness without cause.
But what the integrated figure doesn't take into account is how dynamic your song is. If your song is high energy throughout, with instruments playing from 7 to 8 intensity on a scale of 1-10, the macro-dynamics of your song will naturally be very understated. By comparison, another song that starts out at an intensity of 1 and rises all the way to an intensity of 10 will sound much louder during the loud sections, assuming you set your limiter for each song to target the same integrated LUFS value.
The other approach is to measure using the short-term LUFS value. If the loudest three seconds of your song ring in at -9 short-term LUFS, then your song is likely loud enough to be reasonably competitive while being quiet enough to maintain dynamic range.
Deciding Between Short-Term LUFS and Integrated LUFS
Some mastering engineers recommend using the short-term LUFS value to determine the final loudness of your song. Proponents of this method are right: using integrated LUFS doesn't take into account the differences between a song that's consistently loud and a song that has occasional momentary loudness. In this way, measuring with integrated LUFS can mislead you.
Other mastering engineers recommend using the integrated LUFS value to determine your song's loudness. After all, momentary loudness doesn't define the loudness of the whole song, and this measurement is closer to the algorithms the streaming services use to determine how much to turn down loud songs.
You can use whichever method makes more sense to you. Ian Shepherd prefers to measure by short-term LUFS. I personally find more sense in using integrated LUFS. Either way, you'll have to adjust the volume of the rest of the songs on your album by ear to have them best match the first song's loudness.
The Answer to Loudness Measuring
So there you have it: employ very light limiting at the end of your mastering chain. I recommend setting a target true-peak ceiling of -1.0 dBFS, in part because volume normalization algorithms are sensitive to peaks above -1.0 dBFS and would only turn down the volume of your song, and in part because you don't want the mp3 conversion volume boost to push your song into distortion.
Then place your LUFS-aware metering plugin after your limiter, and play your song from start to finish to discover the peak value of the short-term LUFS and the integrated LUFS. The values may start low and then climb higher during louder portions of your song. That's normal; we're just looking for the final values after your song is finished playing. If you're measuring by short-term LUFS and your metering plugin shows a value of -11.7 after the song is finished, then you can likely crank your limiter up by an extra 2.7 dB (targeting -9 LUFS short-term). And if you're measuring by LUFS-integrated and your metering plugin shows a value of -15.4 dB, you can ratchet up your limiter by an extra 1.4 dB (targeting -14 LUFS-integrated).
Play your song from start to finish again to check that the result of changing your limiter's threshold is as expected and that you met the final LUFS-integrated or LUFS short-term value you were targeting.
Limit any more aggressively than that and the majority of your audience won't hear the volume advantage anyway. Using this method, your music will sound as loud and clear as any on the popular music streaming services, and the dynamics of your music will likely sound superior since your music was mastered in the post-Loudness War era. Congratulations.
You probably need a nap after that. I know I do.
But if you've made it this far, you know which brands make meter plugins that are useful, which loudness scales to pay attention to, and which target loudness is ideal for your music. That's a major win.
Now start releasing dynamic music! And do it without fear. This is the death of the Loudness War, and it's a beautiful thing.