Introduction: What is Loudness, anyways?

If you have spent any time browsing online music production communities, you will probably have come across frequent questions like "how loud should my music be?" and"is my music loud enough?" 

Especially in electronic music, there seems to be great deal of anxiety regarding loudness, with many people worrying that if their music won't reach a certain amount of loudness, it won't be able to "compete".

Far from being a strictly new phenomenon, these kind of competitive anxieties around loudness — what is often referred to as the loudness war — have been around since the dawn of recorded music, when radio stations and record producers discovered that cutting a record louder than others can make it sound more impressive and stand out from the crowd.

However, today it is often said that the loudness war is now "over" — but is this actually true? But before trying to answer that question, is it important to first ask, what is loudness, anyways?

Peak and troughs

The most important concept relating to loudness is the difference between absolute peak loudness (technically, this called amplitude or level, but it is referred to as "peak loudness" here for ease of illustration) and average loudness. The difference between the two is simple, but can be easy to mix up once getting into the nitty-gritty. You often see people mix up average and peak loudness in the context of delivering files for mastering for example, when they take the recommendation of having the final mix render at "-6db" as standing for -6db average loudness rather than -6db peak loudness.

While peak loudness is simply the highest overall measured amplitude of a sound, average loudness is related to the sound's dynamic range — the difference in level between the quietest and loudest moments of the music.

It is often said that the louder (average) a piece of music is, the less dynamic range it has. This is true, but it helps to understand why that is the case for recorded music specifically. Because in the natural acoustic world (measured in SPL, "sound pressure level"), a highly dynamic sound can also have a much higher peak level than a less dynamic sound.

Think of how a static background hum can be very quiet and have low dynamics, whereas something very “dynamic” like construction noises can be dynamic and have much higher peaks than that background hum — dynamics and loudness aren't intrinsically related.

However, that is not the case for recorded music, because unlike the natural world — whose highest possible SPL peak is essentially “unlimited”, since it far exceeds what humans can even physically withstand — there is a fixed arbitrary limit to the maximum amplitude that any recorded sound is faced with.

That limit is the 0db peak point in the digital realm (analog recording mediums like vinyl also have an arbitrary limit, although one that is slightly more variable). While it is technically possible to go over the 0db point, the music will not gain any additional amplitude and the waveform will start to get “clipped off”. 

That is what makes the relationship between the sound's average level to that 0db peak point so important. Because every song is capped out at the same zero db limit as its maximum peak, something that has a lower dynamic range will have more "stuff" in its waveform and thus sound louder. 

When bringing down the peak of the sound (usually the transient) with clipping or saturation/distortion (which can also be forms of clipping), the rest of the sounds is brought up in relation to the 0db peak, thus increasing the average loudness while also reducing the overall dynamic range. This is easy to visually demonstrate in any waveform editor (the first picture is the unprocessed sound, the second the clipped sound):

Measuring loudness: RMS and LUFS

While peak loudness is measured in dbFS (“decibel relative to full scale”), often simply referred to as “peak”, average loudness in music is most commonly measured in either LUFS (“loudness units relative to full scale”) or RMS (“root mean square”). There are more niche loudness measurements like EBU R 128 for TV and film but for music, LUFS and RMS are the most commonly used.

RMS is an old standard going back to the days of fully analog recording, whereas LUFS is a newer standard designed to better capture how humans perceive loudness in music. Since LUFS has become the de facto standard in the modern digital world, in most cases it makes sense to just stick with LUFS. That said, in practice, RMS and LUFS tend to produce somewhat similar, but not identical, numbers. 

Just to name some ballpark real world numbers for LUFS measurements, the vast majority of recorded popular music will sit somewhere in between -5db LUFS and -12db LUFS. The very loudest music (think super loud mainstage EDM) will go up to about -3 or -4db LUFS. Note that at -0db LUFS, the music would have zero dynamic range, because it's average loudness would be the same as the maximum possible peak loudness of zero db, with no room to go up or down.

For the most part, only classical music, jazz and certain kinds of acoustic folk music tend to go much lower than -12 LUFS. That means that most modern music will have a dynamic range of only one or two handfuls of db. And that's fine, for the most part. Contrary to some anti-loudness activists, most genres of popular music do not really want or need a ton of dynamic range. We will get back to that later.

The ear is the best loudness meter

It should also be noted that the implementation of these standards varies, so different LUFS meters by different manufacturers can produce slightly different numbers. It is actually very difficult to find a precise, “objectively” true loudness measurement, because there are different ways of measuring that are equally valid.

For example, the average loudness of music also varies over time — just because the meter reads -8db LUFS at a given point doesn't mean that every second of the song will be exactly that loud. You will get different readings depending on whether you analyze a single bar, a section, or the whole song, which is why there are different loudness metering time frames for LUFS like "momentary" (shortest frame), "short term" (longer frame) and "integrated" (whole song).

This difficulty of measuring loudness also becomes apparent when thinking about average loudness in a real-world production scenario. Say you are clipping or distorting something like a hi-hat from an electronic drum machine. These kinds of sounds often have very high amplitude transient spikes that take up a lot of energy, but are so fast that the human ear only perceives them as a percussive "click" or "snap". You can often clip off several decibels of transient energy without losing much audible benefit while gaining loudness.

But how do you audition this to make sure you aren't accidentally doing any harm to the sound in search of loudness? If you level match between the unclipped and clipped sound with your standard DAW (peak) meter, the clipped sound will suddenly sound much louder since its average loudness has increased, making a comparison of the sound’s actual character difficult.

On the other hand, if you match between the clipped and unclipped sounds by average loudness (LUFS), the clipped sound might suddenly sound weak, because you have removed peak energy without bringing up the rest of the sound to adjust for the lowered overall peak volume.

So even with all the advanced metering tools we have today, it is still often best to match loudness “by ear” — which will always be the most accurate measure of how we actually perceive loudnes in practice.

The “loudness wars” and loudness normalization

As you might already know, all common streaming platforms today employ something called loudness normalization. This means that the platform has an algorithm that measures the music’s average loudness in LUFS and then compares that to a certain target — say, -14 LUFS for Spotify. If the music is louder, it simply turns the file down until it reaches that target, if it is quieter, it turns it up.

There are longstanding rumors that Spotify employs some kind of limiting or compression algorithm, but that is not true — this normalization is a simple clean digital gain change that does not effect the sound of the music at all.

As a result of this normalization, both a quiet jazz record and a super loud EDM record will play at the same target LUFS level, something that wasn’t the case in the days of radio where every song would play with different LUFS level.

But has this actually ended the loudness wars? In a certain sense, since it is no longer imperative to compete for loudness in the way it used to be. But at the same time, if you check the LUFS numbers on any genre that isn't highly dynamic music like jazz or classical, you will find that it will be considerably louder than Spotify's -14db LUFS loudness target.

This confuses many people — wouldn't it make sense to always have your music at -14db LUFS, so that you can have the maximum loudness available, while also retaining the biggest possible dynamic range?

Loudness is not just loudness

To understand why shooting for strict loudness targets is not necessarily a good idea, it is important to consider that loudness is not just about loudness. It's also about density, how tightly you are “packing” the waveform. If you would level match a jazz and an EDM song to the same average loudness, the latter’s waveform would still look more “dense” despite not being strictly “bigger”.

And over the decades, the sound of density has simply become an intrinsic part of the tonal vocabulary of recorded popular music that we tend to associate with terms like “pressure” and “energy” — in most cases, a rap, pop, or techno record just wouldn’t sound right at -14 LUFS.

Actual perceived loudness is also something that is not perfectly captured by LUFS measurements, since human hearing is itself highly nonlinear. Even matched to the same LUFS, something very midrange heavy like an electric guitar solo will always sound louder than a song with a lot of bass frequencies because our ears are more sensitive to midrange frequencies.

And since every song will have a slightly different frequency distribution, it becomes very difficult to determine how loud something will actually sound to human ears by purely looking at numbers. Additionally, the human auditory apparatus is also strongly influenced by sensory cross-modulation (e.g. visual stimuli influencing what we hear) and a variety of psychological biases that are impossible to account for with any meter.

The club question

For those making electronic music, one question that often comes up is “there is loudness normalization on streaming, but what about a club environment, where there is no normalization?” 

While it is true that there is no loudness normalization algorithm in a club environment, there is still human loudness normalization: the DJ!

In a scenario where a DJ would play one of your song next to other songs with higher LUFS levels and wouldn’t account for that at all with the gains on the mixer, that would be less of an indictment of your song than of the DJ — since it is a crucial part of a DJs job to be in control over the overall loudness of the music.

If your track is considerably quieter than others in the genre that could be a problem, both in terms of stylistics (density) and the maximum amount of gain compensation a DJ mixer allows for. But since dance music isn’t highly dynamic music to begin with, in most cases a difference of a few db LUFS in loudness shouldn’t be anything to worry about.

So, how loud should my music be?

There is no single answer to this question, since it is going to depend on the style or genre of music you are making, as well as personal preferences. But for most genres of popular music, there will be a rough ballpark that most songs will fall into.

For most electronic music, anything from -9 to -12 db LUFS would be a good number for a finished, pre-mastering mix. If you compare your mixes to finished songs, keep in mind that those will likely have had a bit of final loudness added at the mastering stage.

Making your music louder

Speaking of the mastering stage, working with a professional mastering engineer can be a great way to give your music the final loudness push it needs. A skilled engineer will be able to push up the average loudness a few db, while also improving the overall mix and presentation. But at the same time, there is only so much an engineer can do at the mastering stage without undergoing serious compromises. A healthy level of loudness always starts from the ground up, at the production level.

In electronic music, one of the most important things to pay attention to in that regard are excess transients. Many drum machines, including classics like the TR-909, have very large transient spikes that can usually be clipped off without losing too much of the sound.

The same also goes for certain synth sounds — especially those coming from analog modular synthesis, which tend to be quite raw, dynamics wise. If you control these spikes from the beginning, the rest of the mixing and mastering process will flow significantly easier.

These days, there are many ways to achieve transient clipping that don’t require any dedicated expensive outboard gear, from clipping and saturation/distortion plugins to various analog mixers, guitar pedals or even a synth's built-in distortion capabilities.

In modular synthesis, many bread-and-butter modules like VCAs, mixers and filters will have analog gain stages that may be overdriven and clipped — this is true for much of Doepfer's line of modules, for example. One of the cool things about using analog-style overdrive and saturation for clipping purposes is that it can add pleasing musical harmonics while also increasing the average loudness, allowing you to "hit two birds with one stone”.

Concerning the question of compression, it is often said that compression “makes things sound louder.” This isn't necessarily wrong, but also not true in every case. For example, If you compress a sound using slower attack times, you can greatly emphasize the initial transient in comparison to the rest of the sound, leading to a decrease in perceived loudness.

Since with the vast majority of electronic sources, there aren’t the kind of natural amplitude variations that would make use of compression for general leveling necessary, compression is usually best thought of as a stylistic choice, rather than as an instant “loudness button”.

While digital brickwall limiters can be a great way to add another couple db LUFS of loudness, they are also notorious for their ability to ruin mixes in a way that cannot be fixed by later processing. A quiet mix can always be made louder later, but an over-limited mix is almost impossible to fix. So it is usually best to leave the limiting to the mastering stage.

The basics

Finally, it is important to remember that loudness perception is always going to be strongly related to the basic fundamentals of music production: are the sounds balanced right, do they fit well together? For example, when the bass of a song is mixed too loud, the other elements will necessarily end up sounding “quiet” in comparison.

These are the kind of creative choices that are not intrinsically about loudness, but will nonetheless heavily influence the perceived loudness of a piece of music, regardless of what its LUFS levels are. In that sense then, it can be helpful to think of loudness as a secondary byproduct of good production and solid music-making fundamentals — frequencies, balancing and dynamics — rather than as an end in itself.

If you are looking for a free and simple loudness meter plugin, Youlean have an excellent free loudness meter for Windows, Mac and iOS.