Loudness Metering
White Paper
Unfortunately, it has become normal and necessary to re-adjust the loudness when “zapping” between channels. The same variance in loudness sometimes occurs at the beginning of TV commercials. Not surprisingly, viewers have complained about these variances in TV programming repeatedly. They are caused by many factors, but there is one common denominator: until today, there is no solution to measure the subjective parameters of loudness perception objectively and precisely. Measuring equipment currently used in audio production does not allow for an assessment of the subjective parameter loudness. Rather, the devices are designed to show the peak voltage of an audio signal, making it impossible to assess the subjective loudness. Quasi Peak Programme Meters (QPPM) are very commonly used among broadcasters. However, with an integration time of 10ms (cf. ARD specifications [Pf3/6]) they are unable to display real peaks. Yet, the detection of these is essential for work involving digital signals.
Paradigm shift in the Audio World
Before we can even start thinking about solving these loudness problems, the way audio levels are viewed needs to changes fundamentally. This includes everything from the currently practiced peak assessment to a uniform loudness measure. Florian Camerer, head of the EBU workgroup P/LOUD, describes the transition from peak normalisation (audio levelling according to peak levels) to loudness normalisation as a fundamental change in the audio world:
»Loudness normalization is a true audio-levelling revolution!« (Camerer, 2010)
Figure 1 shows the two audio levelling paradigms side-by-side. A closer look at the peak normalisation quickly shows that the loudness can vary at uniform peak levels. The audio levelling according to loudness will result in varying peak levels. However, the goal is to keep loudness constant over the duration of the entire program. While this does not imply that differences in loudness in a single program are impossible; the key is the loudness value of the entire program which is determined by a integrated measurement. Loudness and loudness changes are a creative tool when it comes to the design of audio signals. With loudness normalisation, this valuable tool can be used even more dynamically and liberally.
In the past, it was common to work with a certain peak value (generally -9 dBFS in broadcast) that was not to be exceeded. Anything above the peak value was stopped by peak limiters which in turn diminished sound quality. Audio levelling according to a loudness value could very well be compared to “dancing on a string”.
Figure 1: Side-by-side comparison of the audio levelling paradigms (Camerer, 2010)
The one obstacle that stood in the way of the spread of the loudness audio levelling concept for a long time was the lack of an international standard for measuring loudness. Even though a number of methods to measure the subjective parameter loudness had been developed over the last couple of years, none of them was able to establish itself internationally. The situation has changed since then: Since 2006, there is an internationally accepted measuring method. This development coincided with an increased debate about the topic in the audio industry. Already in 2002, the International Telecommunication Union (ITU) had recognized the problem of a nonexistent standard and established a task force to address the issue. The task force was to accumulate the properties necessary for metering methods at digital productions. In this context, they intended to create a procedure with which a reliable indicator for subjective loudness could be established. The resulting procedure is not a proprietary system but an open standard.
ITU-R BS.1770 – Basis of the new loudness movement
The release of the ITU-R BS.1770 marks the first time that an international standard for the assessment of the subjective parameter loudness with an objective, measured value exists. The algorithm for loudness metering consists of a measurement of the energy equivalent average (Leq) with two specific weighting filters. The combination of the two filter curves is called “K-Weighting”. Programme loudness is determined through the voltage sum of the individual channels. In this case, surround channels are rated higher by a 1.41 factor (≈ +1,5 dB). The reason for this is evolutionary, as we perceive sounds coming from behind as louder than those from the front. The algorithm ensures that a 1 kHZ sinus signal in stereo with 0 dBFS leads to a loudness value of 0 LKFS. The ITU Algorithm [BS.1770] shows the loudness values in LKFS (loudness k-weighted related to full scale). Figure 2 shows the design of the algorithm with a multi channel signal. With 5.1 signals, the LFE channel is not included into the measurement.
Figure 2: Flowchart of loudness algorithm according to ITU-R BS.1770-2
In its 2nd revision (ITU-R BS.1770-2), the ITU algorithm requires a so-called gating method. In this context, a differentiation between the absolute and the relative gate must be made. The absolute gate is set at -70 LUFS. Its purpose is to prevent any signals that are inaudible in reality from influencing the measurement. The relative gate is set at -10 LU below a measurement without gate. Its purpose is to prevent that drawn out, silent events falsify the measurement. With the relative gate, levels (measured over a time window of 400 ms) below -10 LU over a long-term measurement without gate are excluded from the calculation.
Beside the method for loudness metering, [BS.1770] describes also a recommendation for measuring peak levels of digital audio signals. The common peak meters cannot display „true“ peak values. The precision of Sample Peak Meters is restricted by the actual samples. Whatever happens within sampling points remains concealed in the audio levelling displays. If a signal includes so-called Inter Sample Peaks – peaks between sample values, following signal processing may cause distortions. For the identification of these peak values, so-called True Peak Meter peak values are displayed after 4x oversampling of the signal.
The ITU standard was seized by transnational organizations like the European Broadcasting Union, and standards and recommendations for use in broadcast were established. Based upon these standards guidelines were developed for the use of loudness measurement in compliance to ITU-R BS.1770 for broadcast.
Standardisation according to EBU R128 and ATSC A/85
The European Broadcasting Union (EBU), an association of 75 broadcasters, has taken up the ITU standard and created guidelines for its uniform use in broadcast. German public television stations ARD and ZDF plan to realize operating principles in accordance with loudness normalisation as proposed by the EBU guidelines by the end of 2011 (Eberhard, 2011). The ITU standard was first applied for the international audio of the Football World Cup 2010 in South Africa; with mixing according to loudness (Krückels, 2010). By now, most manufacturers of audio hardware and metering devices have implemented both the ITU standard and the EBU recommendations.
In 2008, the EBU formed a workgroup under the name P/LOUD with the goal to examine the ITU method for broadcast applications as well as to create specifications that would allow a standardized application in broadcast stations. The team was tasked with the creation of the documents that would serve broadcasters as guidelines for the implementation. The basis hereby is the EBU recommendation [R128]. In four additional documents, details of the implementation of the new method are explained more fully. In [R128], three new parameters that will be used in the future to describe an audio signal are defined:
Programme Loudness
The parameter »Programme Loudness« results from a metering of the signal according to [BS.1770]. This is a long-term measurement covering the entire length of the programme. A specific target level is defined at -23.0 LUFS. In the future, this is the target value for European tv programmes.
Loudness Range
»Loudness Range« is a statistic parameter, describing the distribution of loudness values in a programme. It also allows the description of the dynamic range of a programme. Loudness range can vary from programme to programme and is determined by the content. For this reason, the EBU has not set a specific value for this parameter. Any signal within the loudness range can be adapted to different playback situations with proper dynamic processing. The relevant algorithm can be found in the EBU document [Tech3342].
Maximum True Peak Level
The Max. True Peak Level represents the new, technical top limit of a signal. It was set at 1 dBTP (decibel true peak). This value applies to PCM audio programmes. A lower Max. True Peak Level is required if the signal should be transmitted analog or lossy coded . See the distribution guidelines [Tech3344] for more specific information on the different Max. True Peak Levels.
The EBU document [Tech3341], which provides parameter allowing comparable measurements, describes the so-called »EBU Mode«. Three modes with distinct time scales have been established to measure loudness in EBU mode:
- Momentary-Loudness – 400 ms sliding window
- Short Term-Loudness – 3 s sliding window
- Integrated – start/pause/reset and gating method
The momentary and short term modes are suitable to monitor the current loudness of the signal and can be used to level signals according to loudness. The integrated measurement provides information about the loudness of the entire programme and describes a long-term measurement with the parameters start, pause, and reset.
The absolute unit vary from those stated in [BS.1770]. It describes the absolute value with the unit LKFS. According to international naming conventions, the EBU decided on LUFS as unit for the absolute value. The units for loudness are specified with one decimal and are defined as follows:
absolute: Lk = xx.x LUFS
relative: Lk = xx.x LU
A relative value of 0.0 LU equals the target level of -23.0 LUFS in EBU mode. The results of a loudness metering, for example, can only be defined with the numerical value. If the results of the loudness metering are displayed in a bargraph, the EBU specified the range of the scale. A measuring device operated in EBU mode should offer two switchable scales (see figure 3).
Figure 3: Schematic display of EBU loudness scales (Camerer, 2010)
The section »Practical Guidelines« in the EBU document [Tech3343] contains tips for the practical application of the new method and the use of loudness metering devices.
In the United States, the ATSC A/85 is the applicable recommendation. With the so-called Calm-Act having been passed, it is now embedded in legislation. The recommendation [A/85], published by the Advanced Television Systems Committee (ATSC), is an extensive guideline pertaining to loudness in the broadcast area. It also contains much additional information, such as the proper use of metadata and optimal monitoring conditions. Generally, [A/85] recommends the use of the ITU algorithm without any modifications. Contrary to the EBU [R128], no specific time windows are defined. Only the use of devices compatible with [BS.1770] is recommended. With the target level set at -24 LKFS, the allowed tolerance is ± 2 and the Max. True Peak Level is stated as -2 dBTP. The so-called Calm-Act (Commercial Advertisement Loudness Migration Act) came into effect in December 2010. With the ATSC [A/85] serving as the technical base, the legislation prohibits loudness jumps in television programming now by law. (CBS Interactive Staff, 2010)
Loudness metering in mixing consoles
Lawo introduces loudness metering in mixing consoles for the first time with the production consoles of the mc² series. Scientific research during the product development has shown that loudness metering cannot replace the classic peak meter. Rather, the two tools complement each other.
Nevertheless, loudness metering should not be limited to the final mix, but rather become a tool that is constantly available. For precisely this reason, Lawo has realised loudness metering in each individual mixing console channel. This approach goes beyond simple control according to a loudness value and provides the sound engineer a helpful tool that supports the day-to-day workflow as well as his ears in any situation. The consequent use of loudness metering not only allows getting a grip on loudness problems, but also supports in many other situations, such as the following:
- in the OB-Van or studio with high communication density in the background.
- at additional audio relevant activities that take place during live shows, such as on-tape interviews, delayed recording and transmission, etc
- for naturally occurring tiredness of the ear after long shifts.
A channel based loudness metering or a combination thereof with a true peak meter accordingly allows the creation of a new control standard for digital mixing consoles. An extensive switch to the new control standard will require much personnel training. However, a direct integration into mixing console systems contributes to supporting and asserting this comprehensive change throughout the audio world.
Standards and Norms
[A/85] | Advanced Television Systems Commitee, Inc: Recommended Practice: Techniques for Establishing and Maintaining Audio Loudness for Digital Television. Document A85:2009. Washington D.C., 2009. |
[BS.1770] | ITU-R: Recommendation BS.1770-2. Algorithms to measure audio programme loudness and true-peak audio level. Geneva, 2011. |
[Pf3/6] | ARD/ZDF: Technische Richtlinie 3/6. Aussteuerungsmesser. München, 1998. |
[R128] | EBU: Recommendation R128. Loudness normalisation and permitted maximum level of audio levels. Geneva, 2010. |
[Tech3341] | EBU: Doc. Tech. 3341: Loudness Metering: ‘EBU Mode’ metering to supplement loudness normalisation in accordance with EBU R 128. Geneva, 2010. |
[Tech3342] | EBU: Doc. Tech. 3342: Loudness Range: A descriptor to supplement loudness normalisation in accordance with EBU R 128. Geneva, 2010. |
[Tech3343] | EBU: Doc. Tech. 3343: Practical Guidelines for Production and Implementation in accordance with EBU R128. Geneva, 2011. |
[Tech3344] | EBU: Doc. Tech. 3344: Loudness normalisation in distribution. Geneva, not published yet.
|
Works Cited
Camerer, F. (6. September 2010). On the way to Loudness nirvana - audio levelling with EBU R 128. Abgerufen am 9. Dezember 2010 von tech.ebu.ch/docs/techreview/trev_2010-Q3_loudness_Camerer.pdf
CBS Interactive Staff. (15. Dezember 2010). Obama Signs CALM Act on Volume of TV Commercials. Abgerufen am 15. Februar 2011 von www.cbsnews.com/8301-503544_162-20025823-503544.html
Eberhard, M. (26. Januar 2011). Bericht über aktuellen Stand der 5.1 Ad-hoc-AG. Vortrag beim FSBL-K Workshop "Loudness der neue EBU-Standard" . München.
Krückels, F. (28. November 2010). Roundtable: Mikrofonierung bei Sportproduktionen. Tonmeistertagung 2010 . Leipzig.
