Approaching the nirvana of loudness normalization

RTVE, in collaboration with the AES, has organized a Seminar dedicated to the standardization of loudness and the maximum permitted level of audio signals, debating the application of the EBU/UER 128-2010 recommendation. Continuous complaints from viewers about sound jumps in the audio signal between channels or even within the same channel between the broadcast of programs and advertising or promos are now a tradition. Now, there seems to be some consensus among the industry, broadcasters and advertisers to put an end to this uncomfortable practice.

Jaime Gaiteiro, Corporate General Director of RTVE, Yolanda Marugán, director of the RTVE Institute and Pere Vila, director of Planning and Technological Innovation at RTVE, opened a Seminar this Thursday, February 24 in Madrid, organized by RTVE in collaboration with the AES and with a massive attendance of professionals, dedicated to the standardization of loudness and maximum permitted level of audio signals, debating the application of the EBU/UER recommendation 128-2010.

Continuous complaints from viewers about sound jumps in the audio signal between channels or even within the same channel between the broadcast of programs and advertising or promos are now a tradition. Now, there are numerous working groups that are being promoted around the world to try to advance in an issue that equally affects television stations, operators and all types of production companies. Precisely on January 27, the Secretary of State for Telecommunications and the Information Society hosted the constitutive meeting of the new Working Group on Audio in Digital Television. Its objective is to adapt new audio technologies and promote its quality on television. In addition, the group, coordinated by Pere Vila, will work towards the development of documents and reference guides for the sector on the audio aspects that are analyzed.

Florian Camerer, president of the PLOUD group of the EBU and an expert member of the AES with a long career in the world of audio, has opened the presentation session with a brilliant presentation in which he has made known the current situation and the work that the European Broadcasting Union is carrying out regarding loudness standardization.

Camerer began his speech by stating that “the time has come to undertake a revolution in sound level to solve problems.”

“The EBU/UER recommendation 128-2010 is bringing us closer to the nirvana of loudness to avoid jumps, and now I am very confident that we are getting closer to this paradise of loudness so that viewers do not have to continually use the volume button on their remote,” he stated.

The way in which we are measuring the signals conditions this situation since we take as reference the peaks or, better yet, the almost-peaks (less than 10 ms). These peaks cannot be seen well on the meter, but they are transmitted in violation of the guard margin due to increasing compression in audio sacrificing all dynamic range. Therefore, in Camerer's opinion, compression is not ideal to solve loudness problems with techniques not to differentiate from the competition by introducing more noise but by offering more artistic audio.

Before 2006, there was no international standard on measuring loudness, a key issue since loudness is a completely subjective impression. Finding an algorithm that reflects this subjectivity was not an easy task, although finally the ITU took the first step with the international recommendation ITU-R 1770, introducing the LU (Loundess Unit) and LUFS (Loudness Unit, referenced at full scale) measurements.

Subsequently, the EBU, in addition to the average loudness of a program ('Programme Loudness'), recommended that the descriptors 'Loudness Range' and 'Maximum True Peak Level' be used for the normalization of audio signals, to adjust to the technical limits in the entire signal chain as well as the aesthetic needs of each program/broadcaster depending on the genre and the target audience.

According to Camerer, using normal peak reading, “the QPPM (Quasi-Peak Program Meter) specified in the EBU Tech Doc 3205-E does not reflect the loudness of an audio signal and the QPPM is not designed to measure a long-term average.” Furthermore, measuring program loudness by gate threshold is favorable for improving the loudness agreement of programs with a wide sound range.

For the normalization of loudness and maximum levels allowed from the EBU, the EBU recommends that the descriptors Program Loudness, Loudness range and Maximum True Peak Level be used to characterize an audio signal. On the other hand, consider that the Program Loudness Level is normalized to a Target Level of -23 LUFS. The permitted deviation from the target level should generally not exceed ±1 LU in program types for which exact normalization cannot be achieved in practice, as would be the case with live programs, for example.

Florian Camerer has also emphasized that the audio signal should be measured in its entirety, without emphasizing specific elements such as voice, music or sound effects, carrying out this metric with a loudness meter in accordance with both ITU-R BS.1770 and EBU Tech Doc 3341.

The EBU has also introduced what has been called the gating threshold method by setting a relative threshold of 8 LU below the unmoderated loudness level LUFS specified in EBU Tech Doc 3342, which would be called the “G8 factor” (which will be introduced modified as a G10 value next March in ITU-R BS 1770).

Camerer has stated that “advertising, for example, needs a lively and dynamic sound, so far from resorting to compression, more attention should be paid to the audio design of a production.”

The Austrian ORF where Camerer works is taking measures in this context by setting a maximum range below 20 LU in the broadcast of films and classical music.

Regarding measurement, there are already more than 200 manufacturers that are offering technical solutions in accordance with this new context in accordance with both the ITU-R BS.1770 document and the EBU technical documents 3343 and 3344. While the audio processes, systems and operations with respect to production and implementation must be carried out in accordance with 3343, the audio processes, systems and operations with respect to distribution must be in accordance with 3344 (can be downloaded in here).

Effects on engineering

David Vivas, director técnico y responsable de la oficina de Madrid de Unitecnic, ha hecho un repaso sobre los efectos sobre la ingeniería de las instalaciones que supone la recomendación EBU R128 ya que introduce nuevos términos y nuevos modos en la medición de niveles de audio. La R128 afectaría a toda el flujo de trabajo desde la producción (tanto en móviles como en estudio), la postproducción, el playout (ingesta, emisión…), archivo… afectando en dos puntos: instrumentación de medición en producción, distribución y transmisión, y el cómo y con qué instrumentos realizamos el procesado de la señal de audio. También ha puesto en sobreaviso para que la ahora de adquirir un medidor responda a los requisitos propuestos en la recomendación.

Vivas entiende que “ya que la R128 recomienda el uso de un medidor de loudness en lugar del medidor de pico (QPPM), todos los puntos de monitorado de nivel de audio en la cadena broadcast deberían utilizar medidores de loudness. Esto no quiere decir que se abandone el uso del picómetro para saber qué nivel de decibelios tiene una señal, pero sí debería pasar a un segundo plano”.

En entornos de producción, y dado que cada día se utiliza más audio embebido, además de medir en salida del mezclador de audio y de vídeo, se debería tener en cuenta el nivel de loudness tras pasar por el embebedor de programa antes de la emisión.
In post-production, practically all audio systems are software-based. These systems incorporate or can incorporate software-based measurement tools as functionality of the system itself or as a plug-in: VST, TDM...

As far as ingestion is concerned, Vivas has recalled that R128 recommends working with a program audio level of -23 LUFS, so the contents that reach the ingestion areas should be within this parameter.

For playout, if working with files (ingested or received via network from outside), these may not have the audio normalized to -23 LUFS, this being solved thanks to the existing tools that normalize the audio to the required level by working directly with the files. In broadcast, the entire chain must have systems that allow us to measure the loudness level at each point in the workflow.

Vivas has pointed out that "it would be interesting to use equipment that allows us to record the loudness level of our broadcast over time. This logging data can be crossed with the 'as run log' data of the broadcast automation system to record the loudness level of our broadcast." Regarding the use of dynamic processors in broadcasting, David Vivas recalled that they should be adjusted as a safeguard in accordance with the recommendation.

Los broadcasters

Luis Santillana, head of Audio Frequency Technology at RTVE, has stated that “immersed in HDTV where a greater impact is achieved in communication through a spatial conception of sound, at RTVE we are considering a push for the development and implementation of new features in audio equipment, the management of the 16 channels that allows us to use an SDI frame thus using its full potential, and the non-destructive management of the Dolby E frame so that with fewer channels we can add more information and transport multichannel audio. with specific metadata of the produced program.”

Santillana recalled that since 2008, at RTVE the new measurement environments associated with audio tables in 5.1 production environments have to incorporate loudness measurements. Since the beginning of 2010, procurement files already requested loudness measurement according to the ITU BS 1770/71 standard. On the other hand, the corporation is gradually carrying out an adaptation of the existing systems and tools so that operators can properly interpret the loudness of the programs.

Miguel Ángel de la Torre, sound operator at Antena 3 TV and director of Graudio Institute, has analyzed the role of the sound operator in this changing environment. Considers that all operators should respect the dynamic range (-23 LUFS with a deviation of +/- 1 LU for live programs), also guaranteeing the best signal-to-noise ratio (with a maximum Peak Level of -1db) establishing, on the other hand, a correct signal flow in terms of calculations and processes with floating point digital consoles.

"Loudness can be measured and we can apply different criteria depending on factors such as the predominant content, the time slot or the type of program. To obtain good results we should have factors such as the correct calibration of the studio listening with the appropriate loudness meters in both sound controls, continuities, central control and VTR rooms, post-production, ingest... as well as an optimization of the dynamics control and audio design," highlighted De la Torre.

Tower of Babel

Gabriel Solsona, head of TVE's Sound Unit, has made a historical assessment, recalling that "since the appearance of the VU meter in 1939, its intentional measurement inertia did its best to reflect the perceived loudness of the audio signal. In the 80s gave way to the appearance of the picometer... but all this led to the creation of a Tower of Babel with more and more forms of measurement and hundreds of scales. On the one hand, certain production centers They abandoned the VU meter for the picometer, thus losing the already weak reference they had to standardize the average loudness level. On the other hand, other production centers did not see a picometer until much later, and perhaps the least of them were fortunate to have both meters.

On the other hand, advertising entered the game in order to penetrate the consumer, using all the tools at its disposal and they were none other than the increasingly better elaborated dynamics compressors or even digital processing algorithms called maximizers with fewer artifacts or signal distortion. “This type of compression was not detectable by either the vumeters or the picometers; the first, due to its increasingly obvious distance from what it measures with respect to what we hear and the second, because it is not its mission as a meter,” Solsona stated.

From their point of view, the main difference in audio treatment would not come from genre but from whether a program is post-produced or live. In the case of post-produced ones, the goal is for the program's loudness to be -23 LUFS. To do this, depending on the program being assembled, it is important to ensure that the momentary and short-term measurements are around 23 LUFS, unless you want to emphasize or attenuate a passage. In any case, according to Solsona, at the end of post-production an integrated measurement must be carried out and verify that it is at -23 LUFS. If not, it would be easily solved by attenuating or gaining by an amount equal to the offset presented.

In live shows, it would take a little more practice using the meter and to be somewhat reserved to get the integrated value that runs parallel to the program to end up at -23 LUFs with a tolerance margin of +-1 dB.

Finally, Gabriel Solsona concluded his speech by stating that a global implementation is necessary, since if not, the problem of loudness jumps could worsen.”

Post-production

Carles Vila, technical director at Media Arts Studio, has reviewed the challenges that post-production faces in different areas such as dubbing or advertising. Their first conclusion is that very few clients are aware of the concept of loudness, and even fewer impose regulations on loudness. In dubbing, the level of the dubbing has traditionally been equated to the level of the original version.

To achieve uniformity throughout the process, Vila has pointed out that it is necessary to calibrate the mixing rooms to allow mixing and judging by ear, confirming that the levels correspond to -23 LUPS.

As strategies for loudness normalization, it proposes two alternatives: maintain leveling habits and apply gain or attenuation; or change the loudness normalization without the need for displacement… in both cases to -23 LUFS. For normalization during mixing, we could use dedicated hardware (for example, TC Electronic TM-7) or software integrated into the DAW (such as Nugen VisLM) although in any case the integration must be synchronized with the transport and be a post-fader sample.

With respect to the duality of sound vs. “brickwall” limitation, Vila concluded his presentation by arguing that the limitation to -9 dBFS in a completely digital audio chain is now arbitrary and unnecessary if we adhere to R 128, and inheritance of analog needs. "As recommendation R128 proposes to limit to -1 dBTP preventively, normalizing a source audio to -23 LUFS (without limitation), any subsequent limitation (e.g. to -9) will increase the loudness value depending on how intense the limitation is (limiter threshold). Normalizing again to -23 LUFS post-limitation only implies an unnecessary loss of dynamic margin," he emphasized.

For his part, Nacho Legorburu, director of Audio Spot Sonido Digital, began his presentation by stating that "the audience is tired of being startled and changing the volume of their television. And this situation has to end so that the law of the strongest does not prevail in advertising breaks."

“We have created auditory fatigue and fatigue in zapping… and this is the best recipe to stop watching television and advertising,” Legorburu stated.

He has also reviewed some of the actions developed in Audio Spot to guarantee quality throughout the audio post-production workflow.

This seminar on loudness normalization has concluded with the general request that what today is only a recommendation from the EBU become regulations to protect the viewer from sound aggression and avoid possible disadvantages between broadcasters that adjust their maximum level to the recommended and others that do not.

10 things you need to know about loudness

1. Loudness is the level of sound that people “hear”
It is what we have always called “volume”. Loudness refers to the perceived intensity of a sound piece (music, speech, sound effect...). It depends on the level, frequency, content and duration of the audio, among other things.

2. Viewers and radio listeners complain about loudness jumps
Advertising, in particular, and interstitials, in general, are unfortunately famous for their high level of compression and high level of playback.

3. Current peak sound meters do not reflect loudness
Audio measurement in today's broadcasting is based on peak program measurement (PPMs). Improving sound measurement by adding a loudness meter is one step closer to the best measurement tool: the human ear.

4. Dynamic range has decreased drastically due to the “loudness war”
Peak measurement has advanced the power of dynamic range processing with the aim of making a radio element louder than the competition.

5. Sound dynamics is a creative tool
Dynamic range compression should be used for artistic reasons only. By adopting loudness measurement and normalizing it, audio production can once again look forward to using dynamics as a creative tool.

6. You can put peace to the loudness war by changing the reference
The competition must be in the content and the quality of the sound.

7. Loudness normalization provides the solution
The solution is to change the scheme and mix at a common loudness level.

8. There is a loudness measurement standard: ITU-R BS.1770

9. -23 LUFS, a magic number
This will be the average loudness level of any program, advertisement or interstitial. This is compatible with the freedom of dynamic range as an element of artistic expression. In fact, the maximum peak level goes up to -1 dBTP

10. Normalized loudness must be the standard for the entire broadcasting network
The problem of loudness cannot be solved by broadcasters or equipment manufacturers alone. The normalization of loudness must be addressed by the set of production, continuity, distribution and reproduction activities.