Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/LMJ_a_00976 byguest on 25 September 2021 LEONARDOMUSIC JOURNAL, Vol. 26,pp. 56–59, 2016 56 ABSTRACT state that is known as full scale (FS). Another aspect some- aspect Another (FS). scale full as known is thatstate a 1, to set are byte digital a in bits the all limit—when per vinyl, or a radio transmitter, digitized sound also has an up- one’sandoftapematerial ceiling the toSimilar perspective. race use to suggests [1] Katz Bob specialist mastering to as the referred commonly development is This range. dynamic in trend has led to an increase in electronic levels and a decrease century,this a half perhaps over broadcast, and recording In benchmark. fidelity high a as well as premise gineering en an became and developed were metering, (PPM) peak levels within those boundaries, using both average (VU) and electronicmedium’s optimize to Strategies ceiling. material floor andthe (risk distorting of) the waveform by hittingthe noise the between difference the by defined been has tape magnetic and recording mechanical of range dynamic the Historically,maxima. and minima bylimitations, technical bygoverned are, and been, have levels productionsystems, under water.deep gravity,somewhereof center its by but peak, visible its by defined not is whole, a as which, iceberg, an of that is phor value of the sound wave’s amplitude. A useful pictorial meta- just to values the peak of a waveform but toalso the average Loudness is a perceived quality of sound, which is related not domain ofaudioproduction. of thesedevelopmentsforthenotionagencyinsociotechnical analysisoftheimplications closeswithashort Thearticle at concerts. precedes adiscussionofusingaveragesoundlevelsinmeasurements inbroadcastaudioengineering introduction ofloudnessnormalization ofthe strategies inthecontextoflivemusic.Abriefoverview discussesnewsoundpressurelevel(SPL)measurement This article Average IstheNewLoudest with thisissue. See
loudness loudness - problem of human agency, not one of technology, as becomes ume is better measured in discourse than decibels” [11]. It is a vol- of history “The writing:by paperconcludes his Devine LOUDNE also also Ian Shepherd [8], the initiator of by amistrust of compression [7]. se per louder in general (e.g. Murray Schafer [6]), and exacerbated world gettingour to linked been phenomenonalso The has consumers’ears. to better sounds music louder that ence) experi engineers sound many for (and belief the is there where it competes with ads and songs by others. Additionally cars, (noisy) in playback for optimized is Frequentlymusic producers to be louder than competitors—has several causes. Vickers [3–5], the loudness race—the desire of audio content often mentioned as an example. Fallujah band the with pallet, productionmusic the of part become has trait this styles, metal heavy some forinstance, For [2]. defect a not trait, a are range dynamic marginal a hyper-compressioncontemporarysome styles, musical and for that is rhetoric race loudness the in over passed times parameters. musicalessential assubsequentlyand diminishedsquashed dynamics (e.g. the transients of percussive sounds) are being micro and album) an or track a of range (dynamic macro consequence, a As ceiling. the to closer crept have ducers comenot down; has ceiling Metaphorically:pro-audio The reduced. been haveproduction music in dynamics average able dynamic range (with the move to CD in particular), the avail- in increase anbrought online) CD, tape, cal/electric, However, writes: as Devine There are many examples of activism, notably by Katz but Earl and Kyle Devine instance,by,Well for documented While the historical succession of media (e.g. mechani (e.g. media of succession historical the While most people listen to music [10]. which in situations the in best fare they because prevail sonnel,loudgenerallyandheavilycompressed recordings despite the objections of some musicians and industry per- SS NORMALIZ o:016/M__07 ©2016 ISAST doi:10.1162/LMJ_a_00976 ATI ON dynamic range dynamic
day [9]. - - apparent in how the loudness race is, apparently, coming to the actual output volume of your playback device, which is an end. A few years ago the broadcast community proposed of course up to the user) and storing that setting in the track’s a solution, going back to the notion that loudness is related to metadata. The process does not change the audio, but it de- averages and not to peaks [12]. Prior to broadcast, all (digital) termines a value that constitutes a relative listing level based program material, whether for TV or radio, is analyzed, and on that track’s dynamic properties. The fact that the audio a value for its average loudness is computed and registered in data remains unchanged means purist discussions, such as that program’s metadata. These averages are calculated using with MP3 sound quality [17,18], can be avoided. an audio level weighting based on perceived loudness indi- These developments, flagging the final lap of the loudness cated with the letter K [13], not unlike the familiar dB(A) and race, support what has been argued by Devine: The “history dB(C) weighting curves [14] (Fig. 1). of volume is not simply accretive,” and that the notion of loud Expressed in loudness units (LU), the value (referred to music as “a straightforward reflection of an increasingly loud as LUFS or sometimes LKFS, where FS refers to full scale) world” can be seen in a different light [19]. determines at what level setting a track is broadcast. A loud track, with a reduced dynamic range, will be broadcast at a SOUND LEVEL MANAGEMENT lower level in comparison to a track with a wider dynamic The importance of average loudness is gaining traction in range, i.e. a not-so-loud track. Different elements of program another audio domain. An ongoing challenge for (amplified) material are no longer aligned at their peaks but at their music venues is realizing musically informed sound levels gravitational centers (this is where the image of the iceberg that don’t disturb the neighbors and that take responsibility is helpful again). Loudness normalization is not a scheme of for the sonic well-being of staff, patrons and musicians. The the future; it is currently implemented by many broadcasters commonly recommended occupational health and safety and comes standard with most contemporary broadcasting level of 80 or 85dB(A) exposure over 8 hours is hard to apply and recording technologies. What will take much longer is given that many concerts are louder than that. Muddying this convincing the record production stakeholders to rethink issue is our subjective response to music levels, too loud for overcompressing the dynamic range of their tracks and per- some and not loud enough for others, rarely one size fits all. petuating the loudness race [15]. In one common scenario the issue plays out as a confronta- The terminology in use is revealing: Loudness normaliza- tion between mix engineer (mixer) and system engineer, the tion is replacing peak level normalization. Our perception has latter representing the venue and the former the band. With become the norm, not the material specifics of the distribu- a SPL meter in hand, the mixer is asked to bring down the tion medium of the day. A comparable scheme has been in output level of the mix. The mixer may comply, may pick a use in Apple’s iTunes for some years, known as Sound Check fight or may not be able to do anything short of going on- [16]. It is a good illustration of the asynchronous nature of stage and turning down the guitarists’ amps and sending the the process: When activated, the software scans the audio drummer home. Particularly in smaller venues, simply turn- content in the library, calculating the playback setting (not ing the level down is not always possible. Backline (e.g. guitar
Fig. 1. (left) IEC 61672 A and C weighting curves (B is no longer in use, D is not used for music), and (right) the newer K weighting (ITU BS.1770), sketched in for comparison (thick dashed line). Created by Wikipedia user Lindosland.
Mulder, Average Is the New Loudest 57
Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/LMJ_a_00976 by guest on 25 September 2021 amps) and foldback (monitor) loudspeakers can hamper the concert might exceed the allowed levels in (for instance) 15 Public Address balance directed at the audience and as such minutes’ time, which will allow time to bring down the level act simultaneously as source and noise-source [20]. in a musically appropriate way. In an ideal situation this approach would afford planning HITTING THE CEILING the dynamic range of a concert in advance—asynchronously. One way of dealing with the issue is working with a maxi- Band members (or DJs) and engineer(s) can get together be- mum level, not unlike the ceiling mentioned in broadcast and fore a concert, going over the set list. They can mark which recorded audio. With a set maximum, a device called a limiter song is (usually) the loudest, which songs are quiet, which electronically monitors the sound levels. When the sound songs are likely to be sung along to by the audience (an aspect system’s output goes over a set threshold the level is radically that in some situations can significantly increase SPL), con- and immediately reduced. Driving a system —or to use the structing a map or a cue list that guides sound level norms loudness race metaphor, racing a system —with a limiter will for the duration of the concert. inevitably reduce the dynamic range and average loudness, Considering a concert’s set list as a composition, in combi- making the concert simply as loud as possible, just below the nation with exposure-based rules and measurement systems, set maximum. This has the same consequences that worried can reinforce dynamic range as a parameter in amplified live people in the loudness race: reduction of micro and macro music. Moreover, this approach supports performers in be- dynamic range which, in the case of live music, is correlated coming co-owners of the sound level problem in music venues. with hearing damage risks. This problem is shaped by its many stakeholders, includ- TIME, TECHNOLOGY, DELEGATION ing, in addition to the performers and the audience, venue The technological processes outlined above take parameters staff, neighbors, city councils (who may appreciate a cultural of human loudness perception into account—specifically economy with live music but at the same time need to assure an emphasis on average loudness. With that insight, these that urbanites can sleep without disturbance), hearing dam- developments have the potential to not only end the loudness age prevention, the hearing aid industry, sound engineers, race but also inaugurate a new phase in the area of sound sound hire companies and loudspeaker and paraphernalia level management at concerts. producers [21]. The challenge is to identify practices that In (live) music the cultural, the technological and the can sustain live music without (always) being a nuisance to social converge. This is underlined by the example of decibel neighbors and reduce the exposure to hearing damage risks banking: Rather than replacing human agency, it empowers but don’t (always) impede musical expression. stakeholders to operate in the socio-techno-musical complexity of sound levels at concerts. Dutch philosopher DECIBEL CREDIT René Munnink discusses how the distinction between hu- In recent years the notion of average term exposures (e.g. 15 man and machine emerges from the famous Turing test [26]. minutes or an hour) has been making waves in sound level Munnink argues that the role of arbiter, interrogator and management—not necessarily replacing existing regulations judge is reserved for the human being. The important turn set in instantaneous maxima but to provide stakeholders with with regard to the sound level issues is that a judgment over target values over time in order to work within the prescribed something that is crucially experienced in real time, such as range. One example is found in the Flemish part of Belgium, the loudness of music, can be informed by an asynchronous where recent regulations prescribe a loudness of 100dB(A) process. It detracts from the immediacy of decision-making measured over 15 minutes [22]. In the Netherlands similarly, while providing an option to respond over a period of time not legislation but a covenant was developed between music instead, empowering the stakeholders to find an appropriate venues and festivals, sound hire firms and a hearing protec- compromise. Importantly, there is no influence on the audio tion agency, with approval on a national level for three years by the computer, as was the case with automatic limiting [23]. Currently in the Netherlands the agreed exposure is devices and their technical ceilings. 103dB(A) over 15 minutes. In both countries the new rules A subtle insight emerges from the relation between human also require detailed logging of the measured data. and technological agents. What in Actor-Network Theory The similarity to loudness normalization in broadcast me- (ANT) is referred to as delegation [27] (essentially, the dia is striking. Instead of working with a peak maximum, the transition of tasks from human to machine) can help us to SPL is evaluated over a longer, musically relevant period of understand and evolve in the complex environment of live time, affording a calculated response by specific stakeholders. music. In sound engineering some of these delegations are New software tools with specific interfaces allow working reversible (e.g. the control of levels in a sound recording) but with what is referred to as decibel banking [24]. A set amount other, newer, digital signal processing tasks have no obvious of decibel credit is deposited in an imaginary bank account human equivalent, for instance in the case of real-time pitch at the start of a concert; that credit is exchanged over time at correction. By taking away the immediacy of sound level the rate of the average sound level. This in turn informs the control, new aspects of the socio-technological networks mixer’s decisions, working with the available dynamic range are laid bare. Delegating measurement and computation to to control the sound levels [25]. In the example situation dis- a computer empowers the immediate stakeholders to consult cussed above, the system engineer can alert the mixer that a and strategize.
58 Mulder, Average Is the New Loudest
Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/LMJ_a_00976 by guest on 25 September 2021 References and Notes streaming audio providers such as Spotify and Pandora. At the time of writing, a petition is circulated on change.org with an appeal to 1 B. Katz, Mastering Audio: The Art and the Science(Burlington, MA: that extent. Elsevier, 2007) p. 167. 17 J. Sterne, “The MP3 as Cultural Artifact,”New Media & Society 8(5), 2 Using compression to reduce dynamic range allows for an increase 825–842 (2006). in level, resulting in a louder (i.e. to our perception) track. 18 J. Sterne, MP3: The Meaning of a Format (Durham and London: Duke 3 K. Devine, “Imperfect Sound Forever: Loudness Wars, Listening Univ. Press, 2012). Formations and the History of Sound Reproduction,” Popular Music 32(02), 159–176 (2013). 19 Devine [3] p. 30. 4 E. Vickers, “The Loudness War: Background, Speculation and Rec- 20 In such cases, live pop and rock can be understood as an exception ommendations,” in AES 129th Convention, San Francisco, 2010. to the Shannon-Weaver model—not to mention situations where the audience sings along, in which case the receiver becomes a noise 5 E. Deruty and D. Tardieu, “About Dynamic Processing in Main- source as well. stream Music,” Journal of the Audio Engineering Society 62(1/2), 42–55 (2014). 21 J. Mulder, “Amplified Music and Sound Level Management: A Dis- cussion of Opportunities and Challenges,” Journal of the Audio En- 6 R.M. Schafer, The Soundscape: Our Sonic Environment and the Tun- gineering Society 64(3), 124–131 (2016). ing of the World (Rochester, VT: Destiny Books, 1977) p. xii, p. 301. 22 M. Kok, “Sound Level Measurements & Control at Large Dance 7 J. Corbett, “Free, Single, and Disengaged: Listening Pleasure and the Events,” in Audio Engineering Society Conference: 58th Interna- Popular Music Object,” October 54 (1990) 79–101. tional Conference: Music Induced Hearing Disorders, Aalborg, 2015. 8 I. Shepherd, “Has the Loudness War been won?” Production Advice 23 (In Dutch) M. Van Rijn, “Convenant preventie gehoorschade mu- 2013 [cited 20 December 2015]; available from
Mulder, Average Is the New Loudest 59
Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/LMJ_a_00976 by guest on 25 September 2021