We are in the middle of an era where the need for actual cds is disappearing and the need for more portable musicplayers like Spotify, iTunes and Youtube is growing. Fair enough, it is easier to quickly stream something over WIFI simply it is all within a ‘thumbs’ reach.
As a consequence of that motion something else is happening; we are saying farewell the good old cd-glassmaster which has been a true master for quite a long period of time. But with the coming of all these new (online-) media streaming content a call for a new master is heard.
At the same time the loudness war has started to lose ground. With the coming of the EBU R128 things are changing, levels are changing to be exact. With that norm we can see the return of dynamics again, like there was when VU was still a norm.
We can see changes in televisionaudio already, in most countries you are obliged to adapt if you do not want to risk a fine. But not only tv, also YouTube has changed their audiolevels lately. Not quite to the EBU standards -23 LUFS, but still it is a fair change in level to bring back dynamics.
Investigation of levels
So 2 of the major players in our everyday life have already adapted to the new norm. But is everyone? And with the coming of the new norm, what happens to CD levels? Or at what level do Youtube, Spotify or iTunes deliver content? And what are their standards anyway?
We did a little investigation by comparing levels on a R128 meter of 4 different songs. The sources are YouTube, Spotify web player and iTunes.
The output of the Spotify and Youtube was internally connected through SoundFlower to Nugen VisLM, a well known r128 meter. The tracks from iTunes were bought and measured with the same procedure.
Everything is measured in LUFS (loudness units full scale) to make it easier to translate to dBFS. ‘I’ stands for Integrated, ‘DR’ for dynamic range and TP for peaks in True Peak.
First off, the yellow value with 19,4 stands out, the reason for that is that on YouTube there is no exact same version available of Hardwells Sally. The videoclip contains some quieter passages with nice shots and hardly any audio, but if you would cut out those scenes, you could expect a DR of around 6,5 dB.
The big picture
There are a few things that stand out. First of all it seems that Spotify and iTunes have the same integrated loudness, about -10 LUFS on average (green). And that in itself, is remarkable; it seems it still uses the old norm where the loudness war started from, 0 dBFS.
YouTube seems to use -14 LUFS as a standard (red). EBU R128 says that a 0 LU (the recommended program loudness) is -23 LUFS or dBFS, so all testsubjects are at least 9 dB louder than the EBU recommendation.
Between the three sources from our test there is a 4 dB difference. Mind you, as far as we know now, YouTube themselves adjust the levels of the content, but it sure is something to keep in mind with regards to delivering audio to these broadcasters.
Now iTunes Sound Check advises to use -16.7 LUFS as integrated loudness, but no such level can be seen with the iTunes testsignals, which is a surprise.
We used to have CDs mixed as loud as 0 dBFS and have a RMS of -10dBFS nominal, despite all there is against it. Now, there are people standing up saying that -16LUFS is the new to be standard measured with r128, pretty much the iTunes Sound Check norm.
That would give 6dB extra room dynamics and transients in contrary with what Spotify and iTunes are showing us now.
The tracks on Spotify and Itunes still show heavy clipping, a whopping 5,0 dB even, DJ Hardwell sure has picked the right name for himself! With the new standard of r128 a clip occurs at -1dBFS, so a clip at 5.0 LUFS is 6 dB over. The picture at the beginning of this article is from this track. But clearly this track does not comply to R128 at all.
Looking at the first song, Racoon with Brick by Brick, you see there is no real heavy clipping (if you consider the -1 LUFS rule as precaution).
If only the tracks would have been treated properly as described in R128, they would have seen a True Peak limiter, (also called an InterSample Limiter) instead of a sample peak limiter, like a Waves L3.
With dynamic range, something remarkable can be seen.
Have a look at Bruno Mars dynamic range. Although being louder on Spotify and iTunes, the DR is clearly less, 4,2 dB. But the same song at YouTube is being played with a DR of 9,4. It might be that the master delivered to YouTube really has been mastered properly to try to meet the new standards.
The rest of the songs pretty much stay close regardless what source they are coming from. But none of them has a dynamic range of 10dB or more. Their song in different players seem to have the same masterfile.
The new master
As you see the cd-master as we know it can be layed to rest; there are simply to many differences in standards to be considered a master. The master now should be your 24bit file straight from your DAW. Every file that you derive from that file, must be regarded a copy and adjusted to the asked specification.
Not one of the 3 services tested here do comply completely to the R128 standard.
So what to do with our levels for cd then? Well, I would say please do use a r128 meter in the first place.
If you’re mixing music, use -16 LUFS. -23 LUFS is way to soft for CD and created with another purpose in mind. -16 LUFS gives you plenty of dynamic range. Perhaps your mixes are louder, around -8 to -10 LUFS orso, that comes close to the old CD norm. Then just backup a little, you’ll have less distortion and cleaner transients and even some dynamics back!
When working on a media production for tv or so, you are forced to work at -23LUFS (0LU), and with a given dynamic range from the broadcaster. That just might be a 4dB drop from you regular level, coming from -18dBFS.
Anyway, using the new r128 sure is a nice way to finally get some grip on you total program level of your audio as well as your dynamic range and clipping levels.