I know very little about loudness metering and broadcast requirements, im thinking maybe Boz or Marco would have some good info. Good luck Paul, ill be lurking on this one. I do remember marco doing a video where he compared various levels of loudness for the same track, and demonstrating that Soundcloud put them all at the same level during conversion.

LUFS is RMS level with a much larger time constant.

That's explained in Ian Sheperd's video. There are, however a few minor errors in that video.

The LUFS system was developed by the EBU, not by ITC. It's als a recommendation, not an obligation. ITC has turned it into a recommendation and that has been turned into an obligation by some rather hysterical people on the net.

It's main goals are:

- To keep advertisers from being much louder than everything else.
- To produce a consistent level without hurting dynamic range.

The new compressor hardware that MIGHT be implemented has been developed by the BBC. The idea is to set these compressors to 16 LUFS, for instance. Programs that are louder will be attenuated and programs that are quiter will be amplified. It's far more complex, but it would effectively put an end to the loudness war, because programs at the right level wouldn't be affected and trying to be louder would only result in sounding worse.

The above is meant for broadcast. Voice, music, movies, ads... Primarily for TV, but also for radio. TV because the problem is worse with TV. FM Radio doesn't usually have much difference in level, although some smaller stations might try to modulate beyond the allowed 75%. In DAB you can't go above 100% because it's digital.

Most TV stations in Europe will be implementing the BBC hardware, or have already done that.

The streaming sector is rather vague about how they are implementing LUFS. It looks like they are implementing it in software. Their own software. So you could wonder if it is real LUFS. It's not something to worry about, as they give you an LUFS level to work with. And that's all you need...

So get one of the LUFS VST's and measure your songs. The entire song, as that is what it will be judged by, unless it is a very, very long piece.

That's what I tried, and what totally confused me in how things in the song really change the on screen results. The Cubase LUFS meters give lots of info - but on the few tracks I picked, the song content skews the results. For example, in some of them - and you need to think children screwed up on the floor, pretending to be flowers growing, they start small, than as the music grows, so do they, and when the big last crescendo happens, they shrink back and after maybe a seconds silence, it starts again. This is a loud track at each crescendo, and the music range goes from ppp to fff - with the loudest being around -5dB FS so not even mega loud as a maximum, but LUFS makes the track look on the data to be quiet compared to another, which it isn't! If we were to ever use Spotify or the other LUFS controlled media for distribution, then the carefully planned levels, track to track will be totally wrecked. At the moment, we can happily ignore LUFS, but it clearly doesn't work for our music, and probably most kinds of classical music. I wonder what the BBC do on Radio 3?

LUFS has been designed NOT to compress tracks with lots of dynamics.

I do have experience with classical music, that has lots of range. If the LUFS on the track is around what's prescribed, the compressor doesn't touch it.

I have no experience with what Youtube, Spotify or other streaming services make of it. I also haven't got a clue how these implement it. I do know Youtube changed it's renderer's when LUFS started. And sound quality seems better since, even for things uploaded long before LUFS appeared on the horizon.

As far as I know, LUFS doesn't warranty equal volume perception, specially with dynamic content.
Streaming platform usually scan the whole file and evaluate it's volume as a whole. So, someone with quiet parts and loud parts may end up sounding louder than others.
Honestly I don't know exactly what the algorithm of youtube does and I'm sure youtube doesn't want us to know ..

For me, since I do Pop and Rock music, I use the LUFS mettering of Fabfilter Pro L2 to help me set the limiter to healthy levels (-9 LUFS for a CD)
But I don't analyse all the songs one by one and set them all to -9LUFS.
I do analyse one, and reference the others to it. I also make sure no song gets higher than (-9LUFS) on their louder parts, which in the end often makes them quieter than -9LUFS when analysed completely. So far, I'm getting consistent results this way. But I'm always open to adjust my ways if anyone have better workflow to propose !

There is some great info here, even if you don't have L2 :
https://www.fabfilter.com/help/pro-l/using/loudnessmetering.html

I'm reading all the info and slowly, if I've understood it correctly there is a standard for broadcast that requires -23 LUFS and for Spotify, -14 LUFS. However the maximum for the peaks is -1 dBTP. If this is correct this surely means that Spotify requires limiting and/or compression to meet their spec, and presumably if their algorithm detects you didn't;t squash it, they will squash it for you?

Is this not the Radio 3 vs Classic FM problem all over again but this time for popular music, not classical.

Did somebody think this a good idea? Why the drive for level matching and standardised dynamic range? I spent two hours in a big studio yesterday with the completed tracks from the current project and in the big room, they sound better than they do in our studio - which is good. I tried playing one of our loud and one of our gentle tracks that were tweaked to make the LUFS measurements meet Spotify's requirements and they are horrible, and one, that relies on continuous crescendos, repeated throughout - n' be used at all for it's intended purposes.

Are we happy with specs that only work for certain genres of music? It's possible, perhaps even likely I've just go0t the wrong end of the stick, but I've read BBC, Fabfilter, EBU and other comments now on LUFS and I see nobody jumping up and down being angry, unlike the introduction of MD, or MP3s. Where are all the people complaining their music gets stamped on to make Spotify happy. There's a big difference between the dynamic range of a -23LUFS recording and Spotify's -14LUFS.

Please tell me I totally misunderstand. I really hope I do! Did we really invent a huge digital dynamic range only to be denied access to all of it? What'e the point of increasing dynamic range capability with more bits, only to ban people exploring most of them?

This system has been tested for over a decade. If used right, it caters for all kinds of music and not only music, but everything for broadcast. That's what is was meant for.

If Spotify changes the rules and the compressor, it's not LUFS. And as I said, I don't know what the streamers are doing exactly, as they don't publish their methods.

If the average level is around -23 LUFS your dynamic range should not be affected at all. That's not to say you can't find an example of an edge case that might be amplified or attenuated a tiny bit. But the broadcast compressor as used for FM broadcast wil compress a lot more.

For what the EBU proposes, it's all in the open and can be studied on the EBU website:

https://tech.ebu.ch/loudness/

And there already are a number of open source implementations of the algorithm, like this one:

https://github.com/jiixyj/libebur128

So everyone is free to choose. And that's what the streamers have done. They've made their own implementations.

Hey Paul,
I'm also very interested to learn more about this. Clear guide lines would benifit everyone.
As I said earlier, we don't know what streaming platforms do to our sound. Do they go as far as compression ? I can't tell ..
As far as my youtube channel goes, I know I'm getting louder perceived levels when I use compression on VO content.
Here are the results of 2 uploaded videos :

If someone who reads this knows what « Volume/Normalized 73% / 73% (content loudness -4.7db) means, please shime in !

pcrecord, post: 460517, member: 46460 wrote: Do they go as far as compression ? I can't tell ..

I think there's no doubt they compress. I've only looked at Youtube, but it is clear that a video with really loud noises gets attenuated and a video with very faint sound gets amplified.

I wouldn't know how to extract a compression curve from my tests, as it seems to be a running target. The YT algorithm changes over time and lately, I've been moderatly pleased with it.

As far as my youtube channel goes, I know I'm getting louder perceived levels when I use compression on VO content.
Here are the results of 2 uploaded videos :

If someone who reads this knows what « Volume/Normalized 73% / 73% (content loudness -4.7db) means, please shime in !

I don't know, but I suppose that your VO was at 73% of what they consider max. volume. And that looks reasonable to me.

From what I've heard from people who do video, there is a lot more goin' on. That's why I consider YT unsuitable for judging sound quality of a mic, for instance.

cyrano, post: 460518, member: 51139 wrote: video with really loud noises gets attenuated and a video with very faint sound gets amplified.

Attenuated and amplified doesn't mean compressed if it's done equally to the whole content.. That's where we need more info..
Of course I can make my own tests and get back to this thread to report the results... ;)
I don't really have a problem with Youtube either, but I get that for Paulears it could be troubling when handling classical music...

A simple solution for Paul, is to put your wavs on Dropbox, or one of the other cloud discs..

NO streaming treatment there.

Actually - drop box could be a good delivery medium for this latest project. It's a complete buyout, so we retain the rights, but we're assigning the sale and distribution for the client. It's part of a package, with supplied art work and workbooks etc. We get a commission on units sold, but because we don't trust the end users to not copy them and pass them on willy nilly, we keep the PRS and PPL registrations so if anything gets picked up, we should perhaps benefit, maybe?

For the back catalogue which traditionally has been CDs, and now downloads through our website, things have quietened down - with our customer base satisfied - distribution on Spotify and iTunes could bring in a small extra revenue stream, or trickle - BUT - mangling the music could be a real problem.

I'm feeling dim! I left the Cubase LUFS meter switched on and just carried on as usual, but have noticed that my usual top limit for peaks - about -3 to -6dB typically actually produces me a LUFS figure that sits at close to the requirement for things like iTunes - without actually changing how I mix. I started trying to kind of fight the system, but my mixes seem to be close to the new 'ideal' - so maybe I had it right after all!

That's the beauty of the system, Paul.

If you know what you're doing, it doesn't affect you. But the marketeers who want to be louder than anyone else, will feel the hurt.