Loudness in broadcasting has been a long-standing issue and it is why standards bodies like ATSC and EBU published standards built around BS1770 to help publishers and creators to be able to measure loudness rather than peak audio levels. The EBU R128 standard was launched at IBC 10 years ago and as part of a continuing development of the standards, the EBU has announced EBU R128 s2 to cover streaming services as well as cover a number of other issues. We have the details.

New Tolerances

Originally there were two tolerances: ±1.0 LU (Loudness Unit) for live programmes and ±0.5 LU for everything else. PLOUD chairman Florian Camerer explains…

“It turned out that almost everybody was using ±1.0 LU for all circumstances, but what happened in quality control (QC) sometimes was that an outside production company would submit a programme as -23.0 LUFS but in reality, it was, for example, -22.9. The measuring tolerance defined in 3341 is ±0.1 LU, so that is fine. But then it would be measured by the QC department of the broadcaster on a different but still compliant loudness meter. In this hypothetical case it comes out at -22.8 LUFS, which is still within spec but the QC technician says it is out of spec.”

The tolerances are now defined as ±1 LU for production and ±0.2 LU for QC, which allows ‘potentially accumulating errors’ to be taken into account. A new tolerance has been added to the target for True Peak, which is still -1dBTP (dB True Peak) during production. Leeway of ±0.3dB is now allowed for signals with a bandwidth limited to 20kHz.

We understand that this change to the True Peak tolerance was something the metering and monitoring companies had been pushing for. Florian explains…

“We have all the relevant manufacturers in the PLOUD group and they are very happy about this update. They were always aware of what was happening and gave us feedback. There is no actual need to update equipment but different readings are now better covered.”

New Standard For Streaming

With R128 s2 there is a new supplement dedicated to loudness on streaming services. The AES (Audio Engineering Society) published a set of loudness streaming recommendations back in 2015, with the EBU working on its own approach, for some time as Florian acknowledges, referring to…

“[The R128 supplement 2] has been many years in the making but finally we have reached consensus”.

R128 s2 (Loudness in Streaming) recognises that streamed content - both live programmes and files - has become “an increasingly important distribution method for EBU members”. It points out that conventional broadcasters are now competing with the likes of Facebook, Apple, Amazon, Netflix, Google and Disney, as well as music or other audio-only services such as Spotify, Tidal and podcast platforms, which typically use a higher Target Loudness than is used in broadcasting.

Streaming Should be -23LUFS

While the EBU recommends that streaming output should be produced according to R128 and Tech 3343, with the target of -23.0 LUFS, supplement 2 specifies that the loudness metadata should be used correctly to show the actual Integrated Loudness and “manage the gain and dynamic properties of content”.

The PLOUD group is permitting some concessions for an increased Integrated Loudness for streaming content “if the broadcaster wants to be in control of the quality of the potential dynamic treatment”. In this case, the interim value for the integrated loudness should be -18.0 LUFS with content predominantly containing music permitted to have an integrated loudness of -16LUFS and a Peak-to-Loudness Ratio (PLR) no greater than 15dB.

The new streaming recommendations apply to all distribution paths available to broadcasters. But the tolerances have been brought in partly because of the popularity of streaming to smartphones and similar technology on the move. Florian explains…

“The difficulty is the fact that -23 LUFS is still often too low for mobile devices. The main reason for this is the lower electrical output level - a difference of 10dB - for European Personal Music Players (PMPs) compared to their US counterparts. Fortunately, CENELEC (European Committee for Electrotechnical Standardisation) has updated these specifications so that manufacturers are allowed to have higher output levels again across all countries. That’s great but it will take time so we are allowing -18.0 and -16.0 as intermediate average levels for streaming.”

Florian is at pains to point out that R128 “gives clear guidance” for streaming services, whilst recognizing that some broadcasters are already streaming between -20 and -16 LUFS.

“We still have the clear mantra of -23 but acknowledge the intermediate solution of a higher target level as long as YouTube and others still stream at considerably higher levels, which broadcasters will have to work with if they want to compete with those services.”

There Is Still Work To Do And Improvements To Make

Although Florian is pleased with the improvements in the consumer’s experience using SD services he acknowledges that HD services are still prone to the old problems. Florian’s view is that this is mostly about metadata, particularly if a channel is carrying Dolby Digital or other formats and ignores or abuses the ‘dial-norm’ value (indicating the average loudness of content).

“We have described a comprehensive solution for manufacturers of consumer equipment in Tech 3344 but not too many read that, unfortunately.”

Radio And Cinema

However, that is not the only area that still needs improvement. Florian reports that work continues in the radio sector, while a new supplement covering the potential dynamic treatment of Cinema Content is in the works. We understand that Florian has drafted what will become R128 s3 and is waiting for PLOUD members to respond with comments.

What Do You Think?

There we have it, the next chapter in the development of loudness normalization. What do you think about the move to target loudness for streaming and broadcast to be -23LUFS? Is it still too low or has consumer equipment caught up with the lower level? For me, it’s as much about the dynamic range on both broadcast and streaming that is the issue as I have written about on a number of occasions, and it doesn’t seem that these revisions to R128 address these in any way.