Loudness measurement

It has long been a challenge to achieve consistent sound levels that don't force listeners to readjust their set at every program junction. Today's multichannel world, with playout systems sequencing programs, commercials and trails, means there is no scope for manual adjustment of levels. Instead, volume management policies must be defined. The ITU BS.1770 measurement has been around for a few years and defines an algorithm for loudness measurement. However, it offers no guidance over the values that might offer suitable settings. This challenges manufacturers to keep up with an evolving set of measurement norms by allowing a significant number of user variables. Incorrect setting of these can cause major variations in the values displayed, and a thorough understanding of loudness is necessary to correctly configure a meter so that it gives the required measurements.

As Figure 1 shows, ITU BS.1770 is a helpful start point and defines a number of components, including:

  • a pre-filter giving a 4dB boost above 1kHz to model the assumed spherical shape of the human head;
  • a high-pass RLB (revised low frequency B-weighting) second order filter;
  • an RMS measurement to give values that represent signal power; and
  • summing of multichannel audio, while discarding the content of any LFE channel.

Integration time

Any attempt to make an instantaneous measurement of loudness value is doomed when it is remembered the perception of loudness is influenced not only by amplitude and spectral content, but also by duration of the signal. In effect, loudness has more in common with the “dose” of audio over a period of time, rather than any short-term effect. The time period for this averaging process is a topic that has led to controversy, and different conclusions may be valid in particular situations. For example, measuring short-term segments such as an advertisement that runs for less than a minute makes no sense if averaged over many minutes. Similarly, averaging over a few seconds gives little information about an extended production such as a feature film.

A solution is to assess a number of different values, depending on the nature of the program content. The recently published EBU R128 document and the supplements Tech3341 and Tech3342 call for measurements with a short-term sliding window of 0.4 seconds. This produces an ungated momentary loudness figure, abbreviated “M.” A second sliding window of three seconds gives a short-term loudness, abbreviated “S.” In addition, a third value is an integrated value calculated across a long period such as the full duration of the program. It is abbreviated “I.”

Material with a wide dynamic range will have quiet sections that unless steps were taken, would give an unrealistically low “I” integrated loudness value. Setting a gate threshold can freeze the previous loudness value for periods while the sliding window falls below an agreed threshold.

Getting a good grip of loudness also requires an understanding of the units of measurement that are used. LKFS is the loudness (K weighted) w.r.t. to full scale. This means all units are negative. LUFS is another term having the same meaning as LKFS.

UK and USA legislation

ITU BS.1770 set out the measurement methods, and EBU R128 fixed some of the variables. However, this is far from being a total answer. In the U.S., standards have been defined by the ATSC. Its work has been in part driven by the legal framework created by the Commercial Advertisement Loudness Mitigation (CALM) act.

In the UK, the Broadcast Committee of Advertising Practice, a committee run under the auspices of the Advertising Standards Authority, led the way by establishing some norms for the loudness of commercials. After a consultation process involving the majority of those involved in commercial production, broadcasting and playout, the parties adopted some loudness recommendations in 2008, giving them the force of law through OFCOM — the enforcement body for UK broadcasting.

Using a PPM

This early adoption of loudness standards took place before the EBU and ATSC had defined parameters for loudness measurements. The initial guidance recognized that most organizations had available only the quasi-peak meters that have been the norm in European broadcasting for decades. Interim guidelines, therefore, made recommendations in terms of PPM readings for adverts being lower than that for the surrounding program material.

In the UK, it was initially agreed that acceptable results might be attained if advert peaks were kept to 6dB less than the broadcast peak level below (e.g., PPM 4.5 instead of PPM 6). This gave the UK a head start with a particular advantage for those who used software-based PPMs that could be upgraded to show not only the conventional PPM scales, but could run software that added evolving ITU loudness measurement options.

Initially, there was great uncertainty over sliding window widths, filter curves and reference frequencies, gate thresholds, and absolute and relative levels. For a product to be future-proofed, all of these options had to be adjustable. Equipment could then be configured for whatever loudness standards eventually emerged.

EBU recommendations

The EBU R128 document recommends that all organizations involved should move from the traditional concept of audio measurement on PPMs toward the universal use of loudness meters. It may seem that all the answers are now available, but this overlooks some important shortcomings not yet solved. Loudness management needs to be built into the entire production process. The need is to get it right from the start and not add fixing noncompliant mixes further downstream.

This is critical for today's industry where suitable meters are needed on equipment from the initial recording, through post-production, to the distribution and ingest at the playout servers, and the final delivery path to the viewers and listeners. While much attention was initially focused on television sound, the issues apply equally to radio broadcasting.

EBU R128 recommends adoption of a loudness value of -23LUFS. Extensive tests have shown this can offer a good artistic and engineering solution that meets viewer expectations for consistency of volume. Today's tendency to use music as part of the program style often results in louder sounding mixes. The UK's approach adopted some years ago for commercials worked, and achieved maximum loudness values close to -20LUFS.

Moving to the -23LUFS EBU recommendations might seem simple with a level drop of 3dB. In reality, this is anything but simple, reflecting the fact that commercial broadcasting has to be market driven by airtime sales. That broadcasting is often largely funded by the commercials it shows, and it has to meet the expectations of advertisers who want their messages clearly heard. Having already had a consultation process just a few years ago that resulted in advert loudness being set to a consistent level and extreme compression eliminated, demanding a further 3dB-level reduction would be resisted by the broadcast industry.

As production houses implement managed loudness mixing, they are seriously constrained if they are presented with pre-compressed voice-over and other tracks. The best chance of a good mix, fully compliant with the loudness limits, yet retaining impact, is when any pre-recorded source files do not contain compression. Delivering pre-compressed voice and other tracks limits not only the audio, but also the ability of the commercial production house to mix that material reliably. No specification can be totally foolproof against advanced mixing techniques, and it is to be expected production houses will use their best skills to make mixes that have maximized impact loudness while remaining loudness compliant.

Any real attempt to reduce wear and tear on receiver volume controls requires the entire output of all stations, and not just their ads, to have managed loudness. Many broadcasters use the facilities of playout centers and although the programming is quite distinctive, much of the airtime gets sold to advertisers working across multiple channels. The advert content is common to these channels and it is therefore logical that it is ingested only once, regardless of the channel on which it gets screened. For loudness level to be consistent, the programs between the ads also need a consistent loudness.

If all program and advertisements were created fresh in the light of current loudness norms, all could be well. But, this is not realistic. Much of the material that is broadcast, especially in today's multichannel digital satellite and cable TV world, comes from archive sources. The intrinsic loudness values found in programs created five to 30 years ago varies enormously. The analog origin of much material meant dynamic range was sometimes restricted. Despite this, the styles of mixing in earlier years and the expectations of viewers meant a lot of content was far less-heavily compressed than today. Therefore, it sounds quieter. Today's program content is much more likely to make use of multiband compression techniques that were once the almost exclusive province of music mastering.

Given goodwill on the part of everyone involved and a decision to standardize loudness on some value (whether the -23LUFS of EBU R128 or the -20LUFS used in the UK), all advert and program content could now be made consistent. New tools need to be devised to scan archive material, assess the native loudness it has and then store offset values in the associated metadata that allow playout equipment to send it to air with normalized loudness. There is good progress towards loudness normalization, but much remains to be done if we are to get it from the start to the end.

Karsten Hansen is CEO of DK-Technologies.