The Loudness Wars has invaded the streaming services.

[dasdoing]

What is this LUFS number you refer to and how does it relate to the normal DR numbers we've been using?

'Loudness Units relative to Full Scale' as defined in EBU R128 which references ITU BS.1770 for the measurement method. It isn't measuring the same thing as DR and I don't see how it affects any genre.

since it measures loudness it affects DR. Every genre, generically speaking, shoots for a specific loudness range

 

[danadam]

since it measures loudness it affects DR. Every genre, generically speaking, shoots for a specific loudness range

I suspect they don't mean DR as in dynamic range, but "DR value" as measured by the confusingly named Dynamic Range DB. Changing LUFSi (Integrated) won't change "DR value". "DR value" is more like LRA (Loudness Range), which is different from LUFS.

 

[Sal1950]

The reason I asked and issue to me is, if we constantly jump around using different reference levels and techniques, no one will ever know what the hell the next guy is talking about. I spend a bit of time on the audio forums and consider myself reasonably well informed but don't remember anyone ever talking about LUFSI until it was brought up here today?
I now do understand that LUFSI is completely different than a DR reference but not much if at all different than the Dolby level reference used for most HT setups with Audyssey, Dirac, etc.

 

[danadam]

I spend a bit of time on the audio forums and consider myself reasonably well informed but don't remember anyone ever talking about LUFSI until it was brought up here today?

For LUFSi, maybe, but that's just people not being precise (as usual). Similar to not using suffixes in dB units. But for LUFS there are plenty of hits in the forum search.

 

[Sal1950]

For LUFSi, maybe, but that's just people not being precise (as usual). Similar to not using suffixes in dB units. But for LUFS there are plenty of hits in the forum search.

At my age, maybe it's my memory at fault. LOL

 

[ThatM1key]

Most of (all?) the earlier DACs supporting MQA were decoding it with software running on the XMOS that also acted as the USB interface.

Regardless of how MQA is decoded, it is lossy from start to end.

 

[dasdoing]

normalized integrated LUFS really tell you how much something was limited. dynamic values like LRA are more influenced by composition/arrangement. A song having more quiet passages doesn't indicate quality

 

[RichB]

For LUFSi, maybe, but that's just people not being precise (as usual). Similar to not using suffixes in dB units. But for LUFS there are plenty of hits in the forum search.

It looses data and now money

 

[pkane]

For LUFSi, maybe, but that's just people not being precise (as usual). Similar to not using suffixes in dB units. But for LUFS there are plenty of hits in the forum search.

Integrated loudness, short-term loudness, and momentary loudness are all defined in loudness standards using LUFS (dB). LRA is a dynamic range that is very dependent on the music itself, so only useful if you want to compare two differently mastered versions of the same recording, can't really be used to compare two different pieces of music.

Here's a cool site that lets you measure these: https://lufs.org/

Recent preview version of DeltaWave also performs EBU R128 loudness and range calculations, as well as true peak and plots these over time:

https://www.audiosciencereview.com/...ve-null-comparison-software.6633/post-1777475

 

[IAtaman]

If DR is important to you, to me while secondary to FR it is one of the keys to the "hi-fi" listening experience, streaming is not really a great option on its own. New music is what it is and streaming is fine but for older dynamically mastered music the streaming services generally offer remastered versions which inevitably mean lower DR and sometime radically lower DR than the original versions. There is an easy and inexpensive solution at least for now which is older original CD's. You can pick up thousands of titles for next to nothing ($1 to $2) at thrifts and for a few dollars more pick up rarer titles for $10 or less on the auction sites. That way you can have the dynamic mastering of you favorite older music and not be beholden to the whims of the streaming services, record labels, MQA, etc.

I sense a CD Renaissance brewing!

 

[somebodyelse]

Loudgain can produce most of these too. The docs give some idea of how much of a mess ReplayGain can be with the tag format, let alone differing ways of calculating loudness and different reference levels. By default it uses -18LUFS for consistency with other ReplayGain implementations, but gives a real world example where this results in clipping, and the -23LUFS reference in EBU R128 would be better. Meanwhile I think I read at least one streaming service was using -14 (dBFS? LUFS?) as reference. Maybe that's ok for clipping if they only have more compressed remasters, but it reinforces the lack of a consistent standard @Sal1950 mentioned.

https://github.com/Moonbase59/loudgain#analyze-audio-files-and-output-to-csv

 

[pkane]

Loudgain can produce most of these too. The docs give some idea of how much of a mess ReplayGain can be with the tag format, let alone differing ways of calculating loudness and different reference levels. By default it uses -18LUFS for consistency with other ReplayGain implementations, but gives a real world example where this results in clipping, and the -23LUFS reference in EBU R128 would be better. Meanwhile I think I read at least one streaming service was using -14 (dBFS? LUFS?) as reference. Maybe that's ok for clipping if they only have more compressed remasters, but it reinforces the lack of a consistent standard @Sal1950 mentioned.

https://github.com/Moonbase59/loudgain#analyze-audio-files-and-output-to-csv

Maximum integrated loudness of -14 LUFS is somewhat standard for streaming services, but it does vary:

Spotify      -14 LUFS
Youtube      -14 LUFS
Tidal        -14 LUFS
Amazon Music -14 LUFS
Deezer       -14 LUFS
Apple Music  -16 LUFS

 

[Multicore]

The Amazon Music app has a loudness normalization switch that's on by default. I suppose that's another dynamic effect that's part of the picture here. Or do you think each track has a pre-computed loudness value and the loudness normalization adjusts each track based on that?

 

[danadam]

Meanwhile I think I read at least one streaming service was using -14 (dBFS? LUFS?) as reference. Maybe that's ok for clipping if they only have more compressed remasters,

When a track is below target then youtube will play it as is (it will not make it louder) and spotify in normal and quiet settting will make it louder but only so much so it doesn't clip. I suspect that other services do the same.

 

[danadam]

Or do you think each track has a pre-computed loudness value and the loudness normalization adjusts each track based on that?

Yes. This looks like the simplest solution and I don't see why they would want to complicate their lives with anything else.

 

[levimax]

Yes. This looks like the simplest solution and I don't see why they would want to complicate their lives with anything else.

So does that mean that an album that originally contained "loud" and "soft" tracks will have all the tracks normalized to the same level?

 

[dasdoing]

-14 (dBFS? LUFS?) as reference. Maybe that's ok for clipping if they only have more compressed remasters, but it reinforces the lack of a consistent standard @Sal1950 mentioned.

EDIT: read #119

Youtube for example uses -14LUFS, but if tracks are quieter it wont move them up higher than -1dBFS peak.
Spotify wont move things up ever afaik, unless you set the normalization do "loud" (which is not default)

 

[danadam]

So does that mean that an album that originally contained "loud" and "soft" tracks will have all the tracks normalized to the same level?

Only if you put those tracks in a playlist and listen to the playlist. If you listen to them in the album view, they will retain their relative differences.

EDIT: Also when shuffling the album. From https://support.spotify.com/us/artists/article/loudness-normalization/

• We normalize an entire album at the same time, so gain compensation doesn’t change between tracks. This means the softer tracks are as soft as you intend them to be.
• We adjust individual tracks when shuffling an album or listening to tracks from multiple albums (e.g. listening to a playlist).

 

[danadam]

Youtube for example uses -14LUFS, but if tracks are quieter it wont move them up higher than -1dBFS peak.

That's what spotify does, not youtube. From https://support.spotify.com/us/artists/article/loudness-normalization/

Positive gain is applied to softer masters so the loudness level is -14 dB LUFS. We consider the headroom of the track, and leave 1 dB headroom for lossy encodings to preserve audio quality.
Example: If a track loudness level is -20 dB LUFS, and its True Peak maximum is -5 dB FS, we only lift the track up to -16 dB LUFS.

Here's youtube clip with loudness 15 dB below target:

Loudness: -29.7 LUFS
Sample peak: -10.56 dBFS

and it's played without any modifications:

 

[dasdoing]

That's what spotify does, not youtube. From https://support.spotify.com/us/artists/article/loudness-normalization/


Here's youtube clip with loudness 15 dB below target:

Loudness: -29.7 LUFS
Sample peak: -10.56 dBFS

and it's played without any modifications:

turned it around. will edit post