[MLA] Not either or

[Grimm | Eelco]

Hi Thomas,

I understand what you mean but I am sorry that I do not agree. 

> In the meantime, giving the advice “Album norm only” comes with at least two serious downsides: a) streaming providers might carry only the Album number forward,

That is indeed what is happening at Tidal, since this was the most practical implementation solution for their system, and after discussion with them it did not feel like a bad idea to me. Mark that the majority of albums only have app 3 dB of loudness difference between the softest and loudest track. In my test there were differences of 6 to 10 LU in loudness and some subjects still did not notice it. The reason is simple: with album normalization the vocal levels on all tracks were almost the same (or had a natural balance), the other instruments were responsible for the measured loudness difference. The effect was that in the track normalized version vocal levels were subjectively changing a lot. By using album normalization we introduce the ears of the mastering engineers to balance the relative loudnesses of all tracks. We do not have to rely on an algorithm, which will never be perfect - like we discussed at EBU at great length. Album normalization brings in the human factor into loudness normalization, and by doing so improves the algorithm considerably.

> and b) Album reduces production transparency, a hard-fought feature of BS.1770.

The mastering engineers in this group may need to shine their light on this, but I often heard that track normalization takes away the transparency since loudness differences between tracks that were introduced on purpose during mastering are removed. And when mastering, you know what the louder tracks and the softer tracks will be, so if the loudest track is leading, it is pretty easy to know at what level the other tracks will end. The only exception would be when a soft track is released before the louder tracks. But then still the result is very predictable and therefore transparent: at first the soft track will be the 'loudest' and normalized at -14 LUFS, after the release of the louder track it will be softer, but in the context of the full album that is intended. One could perhaps argue that in the beginning the solitary soft track would play subjectively too loud (at -14 LUFS) in a shuffled playlist with other albums' tracks. But track normalization does not solve this problem. 

Also, mark that there is always a workaround for an artist: just release albums with only one track. That way 'track normalization' can be enforced. The other way around is impossible: you cannot enforce album normalization in a track normalized system, except for releasing the whole album as one track. 

> As a third minor downside, Tidal won’t be able to offer search criteria such as “only suggest tracks with PLR >12 dB” or “tracks with higher PLR at top of list”.

Of course they still can. If the loudest track has a PLR >12 dB, the softer tracks will likely have that too, so the criterium is still very valid. Maybe you would miss the soft tracks of an album that has its loudest track at for instance -8 LUFS. But I don't know if your query is targeted at finding the dynamic tracks of otherwise squashed albums. Besides, Tidal still has the track loudness data available in their database so they can use it for your suggestions if they'd like to.

> Ad a), if the Track value became vital, e.g. because of new regulatory requirements,

Which regulatory requirements are you referring to? Mark that a track will never be louder than the loudest track.

Cheers,
Eelco