The Loudness War
--
Have you ever listened to a song that you really liked, so you felt compelled to turn up the volume of your headphones or speakers? Loud music sounds great, right?
You’re not the only one that thinks so. Mastering engineers (the guys that make final adjustments to songs before being published) have discovered the same thing. So what do they do? they make songs louder in order to make them stand out from others.
But wait.. what if everybody just made their songs louder.. wouldn’t that defeat the purpose?
Yes, and thats exactly what happened in the last 30 years. Once a song was louder than another, mixing and mastering engineers followed the trend by increasing the loudness of their own music as well. This trend has been named ‘The Loudness War’ and it’s still happening today.
It’s naive to believe that there are no consequences to this.
Every war has casualties, right?
So does this one. In this case it’s the consumer of music thats suffering the most. You. Let me explain.
First:
Let’s talk about ‘Loudness’
I’ve been throwing around the term ‘loudness’ a bit loosely until now and fellow producers and audio nerds probably already hate me for this, but let me say:
There’s a lot more to ‘loudness’ than you might think.
Disclaimer: The following sections partially simplify the concepts of signal processing. If you want a more in-depth article about audio concepts, let me know.
Let’s start like this:
To measure the loudness of an audio signal, be it a song or the signal coming from a guitar, a digital or analog level meter is commonly used. They more or less look like this:
This device measures loudness in two ways:
- Peak values
- RMS values
Both of these are commonly represented with the unit dBFS (Decibels relative to full scale), where the loudest possible value is 0 dBFS.
Peak values are an instantaneous measure of a signal’s amplitude. This means that regardless of how long the signal plays, or at what frequency range it is, the magnitude stays unaffected. The upper end of the light green regions on the level meter above show the peak values.
However, the human ear does not perceive sound in a peak-manner.
Instead, RMS (Root Mean Squared) is commonly used to more accurately represent perceived loudness. As the name suggests, RMS is an average of volume over time.
This means that smooth, sustained sounds (e.g.: an openly played note on a bass) have a higher RMS value than short, snappy sounds (e.g.: a snare-drum being hit), even though both might have the same peak value.
So, now that we have the basics explained:
Let’s talk about ‘Clipping’
Let’s say that you’ve made a song and you want to publish it. But shortly before publishing it, you notice that other songs are a lot louder than yours. (A problem that I, myself encountered a lot..)
So, you look at your level meter and see that your song is peaking at around -6dB. So what do you do? Turn up the volume of your song to 0db, right?
Congratulations! You just turned your song’s waveform from this:
..to something more like this:
Where the second signal is obviously a lot louder, as the wave’s amplitude is larger.
|| When looking at waveforms, like the ones above, I think it’s helpful to view them as the speaker cone’s movement over time. If you’re a complete amateur in the field of audio, this might help you visualize this concept.
So, now that your song is 6dB louder, you compare it again with the other tracks out there. Most likely, you will find that your song is still too quiet.
Intuitively, some might think, that we could just increase the volume beyond 0dBFS. So let’s see what happens if we do so:
Now, the waveform exceeds the limits set by the ‘digital ceiling’. However, this cannot happen, as devices cannot output more power than marked by the black lines. Hence, the signal gets cut off, as seen below.
This process is known as ‘clipping’, as the “sticking-out parts” of the waveform are clipped off.
Clipping introduces new harmonics to the original signal. In other words, new frequencies are added, making the signal sound distorted. Distortion makes sounds be perceived as louder.
This should be great! We just found a way to get louder than anybody else!
Well, not exactly. Distortion sounds terrible on whole songs. In fact, audio engineers and producers aim to eliminate distortion as much as possible. Even though it can be a very useful effect, if distortion is overdone, it can easily ruin a piece of music.
This is why clipping is rarely used to increase the perceived loudness of songs.
So, how were mastering engineers able to increase the loudness of their tracks, to a point where we cannot keep up?
Let’s talk about ‘Compression’
Compression is the process of decreasing the difference between the loudest and quietest sound in a recording or signal. This difference is called the ‘dynamic range’.
Simply put: compression makes quiet stuff louder or loud stuff quieter.
Here’s an example:
This is the waveform of an acoustic drum beat:
Now, if we were to apply compression to the signal, the following would happen:
Notice that quiet areas in the recoding received an increase in amplitude. This makes the overall peak level stay the same, but increased the RMS, making the drum beat sound louder.
That’s it. (More or less)
Mastering engineers have been exploiting compression over the last 3 decades, decreasing the dynamic range further and further. This is at the core of the loudness war. This is what the actual war is about:
Who compresses the most. Who gets the loudest song.
But as already said above, also this war has casualties. Compression also comes at a price…
You like Nirvana don’t you?
Well, it’s not bad if you don’t, but it would make the next section a little bit more interesting for you.
Nirvana (that one band from the 90’s) released their legendary album “Nevermind” in 1991. Thats 26 years ago.
Here’s the first song of the album, “Smells like Teen Spirit”:
Looks nice, right? There’s plenty of dynamic range, we haven’t even peaked at 0dB yet and the RMS is hovering at around -12dB. The spikes you see in the waveform, are the transients of the drums. In this song, they’re very ‘punchy’ and clearly hearable. Overall, the song sounds clear and balanced.
|| Transients are very short, loud peaks in a waveform. They can originate from, for example, a drummer hitting a drum. The initial hit is very loud, but then the sound slowly decays over time.
In fact, this record is so good, people still listened to it 20 years later. But obviously, times have changed and the loudness standard has significantly increased. ‘Smells like Teen Spirit’ just wasn’t loud enough anymore!
So what did they do? Remaster the thing.
In 2011, at the 20 year anniversary of ‘Nevermind’, the album was remastered, meaning that the songs were mixed again and more compression was added.
Now, the song looks like this:
Peaking at exactly 0dB and a RMS increased by almost 6dB!
Let me tell you.. it’s loud.
The transients look louder, but actually, they sound washed out and fade into the background of the song. The track sounds loud, but feels a little less alive.
That’s the casualty of the war: sound quality.
And it’s the consumer who has to listen to the compressed music, hence they suffer the most from this.
I’d rather turn up the volume of my speakers, than listen to badly compressed music.
If you want to compare the two by yourself, you can find the original version here and the remastered version here. Note: for an accurate comparison buy the two versions, as the YouTube uploads are not a perfect representation of the songs.
If you think that this is loud, listen up:
Let’s talk about Dance Music
Electronic dance music is known to be louder than most genres. Take this electro song for example:
That’s quite loud. But in electronic music, especially when it comes from experienced producers, music this loud still sounds good. Why, you may ask. Well, that’s for multiple reasons:
- Compressors became more advanced and sound (to a certain extent) ‘transparent’
- Dynamic range does not play a key role in dance music
- Many elements in dance songs are already distorted, making the overall track already sound louder.
But then… there’s Dubstep.
If the music we covered until now, still isn’t loud enough for you, you might wan’t to give Dubstep a shot.
Dubstep is loud. Very loud. But it’s fine.
Loudness is a key part to the Dubstep scene and it has become a stylistic feature as well. Consider this song:
I wish I was kidding, but in the scene of electronic music, this waveform is referred to as a ‘sausage’. With a bit of imagination, it does kind of look like it.
What should I say? This is loud. Really loud. The peak AND the RMS are at almost 0dB. Practically speaking: songs can’t get louder than this.
So what’s next? Are all songs going to look like a ‘sausage’ one day?
I don’t think so. People like dynamic range more than loudness. At the end of the day, the consumer defines the product and I think it’s the same in the music industry. When music continues to get louder, there will be a point where compressors cannot compress more and at that point, the market will turn around and seek the much wanted dynamic range again.
From an artists point of view, I find it very difficult to get my music on the level of loudness set by professionals, while still sounding good. I do respect the skill needed to bring songs to an RMS of -3db, while keeping a clean song. Maybe I didn’t specify enough how compressors work and how difficult it is to make them sound great, but if you wan’t to know more about them, I’d happily write an article on that. Just let me know.
Feel free to give your opinion on this article and let me know if I missed anything. Also, if you’d like me to talk about something, feel free to ask.
-Nico
