Loudness Normalization for Audio Programs in SL NEO Servers According to ITU-R BS.1770, EBU R128

Note: the function is available starting from the version 2.0.26

This function is optional, its activation requires additional licensing.

In the process of development of the module SL NEO Loudness Normalization, tests and control measurements were being implemented with using software of the company NUGEN Audio, UK.



Problem Description

Lately (especially in the century of TV advertising), the necessity of volume leveling of TV programs has acquired special actuality, as the volume differences between the on-air fragments is essentially hindering the viewer to comfortably perceive the broadcasting content.


Till recently, appraisal of the audio level had been being realized (and is realized till now) on the basis of the measurers readings that register maximal or average values of the electrical level an audiosignal, which is not quite correct in modern conditions.


Historically it happened that for volume leveling, the broadcast and production complexes mostly use VU-meters (Volume Units meter) that have the integration (the measuring range) time about 300 msec and belong to the measurers of the level of “average” time values of electrical levels. These devices are applied quite successfully, in spite of the fact that they don’t react to the short impulses (peaks) and don’t display the true volume level.


Although, the use of QPPM (peak) indicators with the integration time of 5 msec displays a true picture of the instantaneous values of the level, the evolution of audio production makes it less reasoned comparing to VU-meters and leads to even bigger volume differences between on-air fragments with different dynamic ranges.   


As it turned out, the problem of volume leveling in broadcasts cannot be solved on the only basis of measuring the values of the electrical level of an audiosignal with the VU- or QPPM-meters and normalization of levels by peaks, so as the subjective perception of the sound “loudness” by the listener is not always explicitly defined by the value of “electrical” level of the audiosignal.


The picture illustrates the reason, why the volume level (the analogue of the sound pressure), actually perceptible by the listener, is not comparable with the “electrical” levels of audio signals which are kind of the same for both adjacent programs. Even with the identical “peak” or “average” level of the audio signal the volume indicator differs. The main reason is the active use of the algorithms of compression of the dynamic range during the production of “short” clips - commercials and music clips. Herewith, active usage of such processing tools is not acceptable because of worsening the quality of the sound picture, which causes problems of volume differences between the commercials and fragments of movies or programs.  




According to the EBU R128 recommendation, the following parameters are applied for the appraisal of the volume parameters, equivalent to the sound pressure:

·         Programme Loudness;

·         Momentary Loudness;

·         Short-Term Loudness;

·         Loudness Range;

·         Maximum Permitted True Peak Level.

The parameter “Programme Loudness” should be normalized to the level -23.0 LUFS, with the permissible variation ±0.5 LU. 


In case, when the exact normalization of the program volume level to the target level is practically unattainable (for example, during “live” broadcasting), the variation of ±1.0 LU from the value of target level is permitted.


The program volume level must be measured as a whole, without selecting out specific elements, such as speech, music or sound effects.


The parameters Momentary Loudness and Short-Term Loudness are used as an addition to the parameter Programme Loudness for the control of current dynamics of the program.


Loudness Range is measured in LU, showing the dynamic range of the program volume.

Implementing the function of normalization of the volume level in SL NEO platform server software.

Functionality of SL NEO server software allows users to normalize not only the “electrical” level of the output audio signal (DRC function), but also the volume level Programme Loudness of the audio-program formed by the server.


Sources for the program formed by the server are the file materials and Live-sources - the signals that come to the server inputs and are switched by the server into the output program according to the events in playlists.


The given function of normalization is actual, when forming SDI-signals with the embedded sound, SDI/analogue video and analogue audio, but it is especially actual when forming IP/ASI transport streams, so as in the last case expediency of installation of external hardware devices for decoding and further encoding to TS only for audio processing is quite doubtful.


The principle of work of the module Loudness is concluded in measurement of the volume level of the input signal coming to its input, according to the algorithms ITU-R BS.1770, EBU R128 and smooth changing the gain coefficient for bringing the output parameters of the volume level in correspondence with the set values.


Very conditionally, the algorithm of the module work may be compared with the work of a sound engineer, who periodically moves the fader on the sound controller, depending on the input volume level, for maintaining the average specified volume level of the output program.


Let’s consider the details of configuring the functional for normalization the volume level at the SL NEO server platform.

Adjusting the volume level of the output audio program

Configuring the parameters is implemented from the control panel - Administrator Control Panel. Login to the console is implemented locally from the server, or from any machine in the network by the address http://server_ip:7901. You should login to the control console as an administrator.


After the login to the control console, choose Manage at the console menu. Choose the tab Video IO Boards, then move to editing the parameters of the appropriate Playout service. Activate the function EBU-R128 in the item EBU-R128 Loudness Enabled of the tab EBU-R128 Loudness.




In the field Loudness Level LUFS you should set the needed reference value of the volume level of the audio program Programme Loudness - minus 23.0 LUFS, in accordance with the requirements EBU-R128.


In the field Integration Period specify the value for the time of measuring the volume of the audio program Short-Term Loudness that comes to the module Loudness, in seconds (in most cases, the input of the service Playout and the module Loudness respectively receive the signal from the service for playlist execution Program Channel). The parameter Integration Period will affect the time of reaction for the change of the output level. The optimal value is 2...4 seconds.


In the field Maximum Gain set the maximum value of the gain coefficient of the module Loudness. This restriction is necessary in case, when the program has the pause in sound and the module gain will be automatically increasing. The measurement units - times, the optimal value is 1.2...1.5 (1.6-3.5 dB). In case of setting the value equal to one, the module will be only weakening the input signals. 


In the field Gain Adjust Time set the value for the time of reaction of the module to the changes of the output level, in seconds. The optimal value is 0.2...1. The actual time of reaction will be bigger, considering the value Integration Period. 


In the field Integrate Channels you should choose the audio tracks, to which you will apply the measurement algorithms. Change of levels will be applied to all tracks, involved in the system.  


In the presence of “technical” channels, channels of multilanguage post scoring etc., it is not always right to measure volume by automatic summing all the channels. That’s why the given mask allows users to count the loudness levels by the chosen number of audio tracks. 

loudness2.jpgControl of LUFS levels

SL NEO server software allows users to dynamically control the Short-Term Loudness level of the audio program in the LUFS format, that comes to the module Loudness, to appraise the current gain coefficient and dynamically control the output loudness level in the LUFS format.


Control of the parameters is implemented from the control panel - Administrator Control Panel. Access to the control panel of the server by the network: http://server_ip:7901.

After the login to the Administrator Control Panel choose the tab Status. The appropriate window Playout dynamically displays the current values of the gain coefficient Gain (times), the values of the volume level Short-Term Loudness on the input and on the output (Before, After) in LUFS units.

Static Change of the Output Sound Level and LUFS Level respectively

If needed, you can statically increase or decrease the level of the output audio signal and the volume level respectively. Adjusting, unhurried configuring in dB is implemented in the menu of the control console: section Manage, tab Video IO Boards, item General, field Base Audio Level.


Installation should be implemented experimentally, with using external certified indicators. This procedure may be actual in case of digital-analogue conversions of the audio signal and mismatches in standards in the area of analogue reference electrical levels.

Recommendations for improving quality of the formed audio program, concerning exclusion of artifacts when automatic normalization of the volume level.

1. Stick to the single average corporate VU audio level in media files formed in the production systems and received from the content suppliers (for example - 12 dB FS conditionally), avoid the difference of adjusting levels between the programs/files. Import files to the server with the identical level audio parameters. 


This circumstance is connected with the fact that the module Loudness is not able to instantaneously react to the change of the signal level and the volume level respectively in the clips that are played back end-to-end in the playlist. In case of discrepancy of the middle levels of adjacent clips, there is a possibility of considerable “splashes” on the borders, which it is impossible to compensate. Discrepancy of the dynamic ranges in the adjacent clips will only exacerbate the situation.


audio_norm.jpg2. Use the built-in tool SL NEO software for normalization of audio levels during the manual and automated import of files onto the server array, if the initial audio levels are not known.


The optimal parameters for import of files with normalization are: Value: -12 dB, Peak detection interval: 200...300 ms.


3. If possible, bring the levels of the external Live-signals sent to the server inputs, to the same “electrical level” readings that the files have (for example, by the VU scale). The less will be the differences of levels during transitions between events - the more comfortable they will be for the audience, the less noticeable will be the work of the Loudness module while changing the gain coefficient in the process of normalization and forming the final volume level.

Test files for checking the correctness of settings for the module Loudness Normalization

Tone 1000 Hz, with the level -23 dBFS, which conforms the volume level -23 LUFS on the given frequency, the duration is 20 sec


Tone 1000 Hz, with the levels -25,-20,-15,-10,-5 dBFS. The duration of every fragment is 20 sec.


The picture displays the diagrams of levels in the test file and the result of work of the module Loudness with the diagram of the transition processes. After changing the input level, after some time of reaction (2...4 sec), the module smoothly changes the gain coefficient and brings the output volume level to the accordance with the set value, providing in such a way the auto-adjustment of volume.


In case, if automatic change of levels is impossible for some reasons, you should use the function of preliminary normalization of volume to -23 LUFS for every file (the functional will be available in the SL NEO platform software in the near future, you can use third-party software for testing - for example, the product of NUGEN Audio company).


Nevertheless, the function of auto-adjustment may turn out to be necessary when working with Live-sources - the signals that come to the server inputs.