The evolution of audio loudness and mastering in the Internet age28 June 2016
Jon Schorah, founder and creative director of NUGEN Audio (pictured), looks at the evolution of audio loudness for music mastering, particularly as the requirement for loudness normalisation has spread to Internet-based streaming services
While many audio professionals have traditionally taken the view that “louder is better,” the emergence of new paradigms for delivering and consuming music is drastically changing the rules for many situations.
Making a song louder is achieved by a technique called compression. This works by reducing the dynamics in the recording (the difference between loud and soft) and then turning up the resulting audio by the amount it was reduced.
Compression affects the sound in two ways: it’s louder, and it’s less dynamic. Louder audio always sounds better (up to a point!) and reducing the dynamics adds cohesion, but it can also destroy transients, or high-amplitude, short-duration sounds at the beginning of waveforms. If overdone, removing transients can drain the life and sparkle from a sound, leaving it relatively dull and lifeless.
Compression techniques have become very advanced, so much so that we are now in an era of “hyper-compression” in which compression techniques are significantly affecting the sound of the audio, as well as making it louder.
When louder actually is better
In a peak-normalised environment, there is a perceived advantage to loudness. For many years, loudness was a commercial imperative based on the idea that if peak-normalised music is louder, it sounds better and stands out more from the music around it. In fact, that concept formed the basis for the “loudness wars” dictating that every mix should be as loud as possible. Put another way, in the peak-normalised world music that is louder and stands out more is preferable to a quieter, more dynamic sound with lively transients.
Loudness is also an advantage in analogue radio, since the greater the energy in an analogue signal, the further it will travel. And the further the signal travels, the wider the audience and the greater potential for advertising revenue; hence the commercial imperative for a radio station to be as loud as possible.
There is also a popular argument that some genres of music, such as the “club sound,” are by their very definition meant to be loud. The argument is that hyper-compression techniques are an integral part of the experience and loudness is desirable in and of itself.
For example, the original recording (pictured right) of ZZ Top’s “Sharp Dressed Man” in 1983 (left) and in 2008 (right) demonstrates how the use of powerful compression and limiting techniques introduced a significant gain in overall loudness (at the expense of transients and overall dynamic freedom).
Enter loudness normalisation
Recently, however, we have seen sweeping changes in the way music is delivered and consumed, and the rules are changing for music producers on how to mix and master content. As the loudness wars have grown fiercer in recent years, they have been fought more ferociously in some genres than others and the intention and skill level of mix practitioners has varied widely. The resultant variation in loudness from track to track has not translated well to streaming services, forcing consumers to reach constantly for the level control.
Radio broadcasting has also changed with the advent of services such as DAB. Since DAB is a digital delivery system, there’s no longer a competitive advantage in transmitting louder audio in order to achieve greater coverage – the signal is all 1s and 0s and covers the same distance regardless of the music’s loudness.
To overcome these issues and deliver a more satisfying musical experience for customers, content providers have introduced loudness-normalised playout. Apple’s iTunes Radio, for instance, is regulated by the Apple SoundCheck algorithm as specified by the internationally recognised ITU-R B.S. 1770 loudness standard, the same standard now being used around the world for broadcast. Following Apple’s lead, Spotify, YouTube, Pandora, MS Windows, music TV stations and, increasingly, DAB radio are also employing loudness normalisation techniques to ensure that the customer listening level remains consistent and satisfying from song to song.
In short, loudness normalisation checks the music for its average loudness level and then adjusts it up or down, so that the entire audio stream is made up of music that is at the same average loudness level. The so-called “loudness advantage” disappears since the playout system removes any loudness introduced over the target value and simply turns loud music down to match the context.
As a result, the use of audio compression in music production returns to its roots as a creative decision, rather than a commercially driven imperative. With a streaming service, it is impossible for a track to “be louder” than others; therefore nothing is gained from compression unless the compression effect itself is the goal (as in club music). Producers that need more transient control can apply compression to their taste, since the loudness will be removed downstream by the playout service. However, hyper-compressed audio will generally sound duller with a streaming service than audio using more dynamics with a greater peak-to-loudness (PLR) ratio.
Again, the original recording of ZZ Top’s “Sharp Dressed Man” in 1983 (left) and in 2008 (right) after being loudness-normalised to -16 LUFS (pictured right). In a modern playout situation employing the Internet “standard” level of -16 LKFS, the loudness advantage is removed, leaving the overall loudness perception the same. Here, the 2008 version clearly lacks transient detail and fails to exploit all available digital headroom.
Mastering tools for the internet age
In order to produce for these new playout environments, it’s important to understand the idea of “loudness introduced over the target value.” Playout systems have an average loudness target around which all audio is balanced. Engineers need tools that can not only measure target levels but help them understand the audio dynamics in more detail. These considerations were secondary in the old days when everything was hyper-compressed, but they are now front-and-centre. In other words, if there is no point in being louder than the target playout level and if engineers wish to maximise the available dynamic headroom, they need built-for-purpose meters that can tell them how they’re doing.
NUGEN Audio has addressed these considerations with MasterCheck (pictured above), the first music-industry-specific audio plug-in designed to facilitate mix and mastering for loudness-normalized playout. With MasterCheck, producers are able to mix audio according to the new loudness criteria and understand not only how algorithms from the major streaming platforms will affect the mix, but how the music will sound on consumer playout devices for the optimal listening experience. The tool enables them to evaluate audio levels against those of target playout systems and international standards, and they can gauge just how dynamic a mix is using the built-in PLR metering. Plus, MasterCheck’s broadcast quality true-peak metering allows the engineer to guard against inter-sample clipping and enables solid translation of mixes to compressed audio formats such as mp3 and AAC.
In addition, modern audio delivery is predominately digital and often employs data compression codecs to facilitate economical streaming and storage. These codecs struggle to handle level overload, which can result in distortion, for example mp3 “fizzing”. Therefore, the ITU has developed a new, more sophisticated measure of digital levels known as true-peak level. True-peak limiting with a tool such as NUGEN Audio’s ISL (pictured) can be used to protect against codec overload, delivering a mix to which data compression codecs can be safely applied without introducing significant distortion.