You are here

Two songs with exactly the same LUFS but one still sounds louder?!

For everything after the recording stage: hardware/software and how you use it.

Moderator: Moderators

Two songs with exactly the same LUFS but one still sounds louder?!

Postby DonGoliath » Fri Nov 13, 2020 9:11 am

as we all know streaming services are using loudness normalisation these days. but it doesn't really work since when listening to music on youtube/spotify i still hear a big (perceived) loudness difference amongst songs. eg. check these two songs having exactly the same LUFS (you can check a song's LUFS on youtube by right clicking in the vid and then chosing "stats for nerds" and then checking the volume/normalized section):

Song 1: https://www.youtube.com/watch?v=P_P1wKYl0cY

Song 2: https://www.youtube.com/watch?v=hRgcgcTP7nM

Song 2 sounds much louder to me. Now i have two questions for you all:

1. Why does song 1 sound louder than song 2? I suspect it has to do with their the tonal balance (with song 2 having more top end which is perceived as louder) and/or the transients (hence the drums) being mixed more upfront? It can't be the dynamics of the songs since i especially chose an example with 2 songs having the same LUFS value.

2. Do you agree with me that it is nonsense that the streaming services use loudness normalisation since even the new LUFS measurement system obviously doesnt work since there are still big differences in perceived loudness? I am wondering why no measurement system can be build to sort this issue. I mean LUFS actually was built to take into account criteria such as tonal balance (outdated fletcher munson curve or uptodate ISO 226) as well as dynamics (measuring already better than RMS) but still LUFS fails to measure how our ears perceive loudness. Can't this be done properly? And if it can't what's the point of using loudness normalisation anyways?
User avatar
DonGoliath
Poster
Posts: 18
Joined: Wed Mar 09, 2011 1:00 am

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby awjoe » Fri Nov 13, 2020 9:58 am

Cuz it works better than what went before?

I'm still going to follow this thread avidly.

https://www.soundonsound.com/forum/view ... 16&t=74985
User avatar
awjoe
Frequent Poster (Level2)
Posts: 2990
Joined: Tue Mar 08, 2011 1:00 am
Well, maybe I could take one for the team. But no pictures.

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby RichardT » Fri Nov 13, 2020 10:25 am

I agree with you that songs with the same LUFS do not always sound equally loud. I’ve no idea why this is, or whether this is a design feature or a flaw. I guess it might relate to changes in dynamics: if a song has uniform dynamics it might sound less loud than one that has big dynamic changes.
RichardT
Frequent Poster
Posts: 749
Joined: Fri Aug 13, 2004 12:00 am
Location: London UK

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby Matt Houghton » Fri Nov 13, 2020 11:20 am

Well, long story short, Song 2 (Sub Focus) is better mixed and better arranged, with more thoughtful instrumentation.

It's cleaner, with less masking going on generally. It has more macro dynamics built into the arrangement, which means that the big bits can be louder without taking the integrated LUFS too high. And micro dynamics too — lots more loud/quiet contrast due to the shorter, punchier sounds used. So again, these sounds can be louder than an equivalent sound that lasts longer to make the same contribution to the integrated LUFS.

Compare and contrast the drums in the two songs, for example. Song two has punchier sounds and more of the groove elements panned straight down the middle where things sound more solid. The drums in the first one (Don Goliath) sound weirdly diffuse to me.

Song two also makes more effective use (I'd say 'better', but it's a question of taste!) of other ear-grabbing tricks, such as the long drawn out pitch rises and so on.

That's not everything to note, but it's already a lot! Seems to me loudness normalisation is working pretty well — in that the better mix sounds better.
Matt Houghton
Frequent Poster
Posts: 1074
Joined: Wed Aug 08, 2007 12:00 am
SOS Reviews Editor

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby CS70 » Fri Nov 13, 2020 11:26 am

DonGoliath wrote:as we all know streaming services are using loudness normalisation these days. but it doesn't really work since when listening to music on youtube/spotify i still hear a big (perceived) loudness difference amongst songs. eg. check these two songs having exactly the same LUFS (you can check a song's LUFS on youtube by right clicking in the vid and then chosing "stats for nerds" and then checking the volume/normalized section):

Song 1: https://www.youtube.com/watch?v=P_P1wKYl0cY

Song 2: https://www.youtube.com/watch?v=hRgcgcTP7nM

Song 2 sounds much louder to me. Now i have two questions for you all:

1. Why does song 1 sound louder than song 2? I suspect it has to do with their the tonal balance (with song 2 having more top end which is perceived as louder) and/or the transients (hence the drums) being mixed more upfront? It can't be the dynamics of the songs since i especially chose an example with 2 songs having the same LUFS value.


Ah! I've just finished a small piece on exactly the subject, after re-writing it three times to get an angle which was both informative and short (hopefully)

Here it is: https://www.theaudioblog.org/post/same- ... t-loudness

2. Do you agree with me that it is nonsense that the streaming services use loudness normalisation since even the new LUFS measurement system obviously doesnt work since there are still big differences in perceived loudness?


Not at all. :) Dynamics are GOOD! The point of normalization is not to make everything flat and boring, but to give control of the overall level back to the user, rather than the mastering engineer. And a track with better dynamics will show off with respect to a less interesting one.

The user has a volume control, which he or she will use at will.

I am wondering why no measurement system can be build to sort this issue. I mean LUFS actually was built to take into account criteria such as tonal balance (outdated fletcher munson curve or uptodate ISO 226) as well as dynamics (measuring already better than RMS) but still LUFS fails to measure how our ears perceive loudness. Can't this be done properly? And if it can't what's the point of using loudness normalisation anyways?

Well, it's probably possible to improve the perceptual model, but in the end of the day no two people perceive exactly the same, so it's impossible to satisfy everyone. Most songs in the same genre will have a chance to sound similarly loud.. if well engineered and mixed.

Good songs, arrangement and mixing shine off more, not less.

The point of LN is to give the control of "overall energy in the room" to the listener. But different songs stay.. well, different. :)
User avatar
CS70
Jedi Poster
Posts: 7738
Joined: Mon Nov 26, 2012 1:00 am
Location: Oslo, Norway
Silver Spoon - Check out our latest video and the FB page

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby Hugh Robjohns » Fri Nov 13, 2020 11:34 am

DonGoliath wrote:... when listening to music on youtube/spotify i still hear a big (perceived) loudness difference amongst songs. eg. check these two songs...

Listening on little laptop speakers I don't hear a huge difference in loudness between those two at all, and I certainly didn't feel the need to change my listening volume, which is the raison d'etre of loudness normalisation. So I'd say it was actually working pretty well...

However, I agree that song 2 sounds better to me, and is more involving and exciting, and that's probably because it makes far better use of micro dynamics inthe mix and arrangement, giving the impression of energy without actually sounding louder overall.

It may also be that the two tracks appear to have different perceived loudness when auditioned on a full range system because of the low end balance.... I'll check later if I remember!

I suspect it has to do with their the tonal balance (with song 2 having more top end which is perceived as louder) and/or the transients (hence the drums) being mixed more upfront? It can't be the dynamics of the songs since i especially chose an example with 2 songs having the same LUFS value.

The internal micro-dynamics of a song can certainly affect the perception of loudness more than the integrated loudness measurement. This is reflected to some extent in the Short-term and Momentary loudness measurements which aren't currently taken into account in calculating the Integrated Loudness value that determines the overall playback level in most streaming systems.

However, this is a known issue, and the Short-term measurement is being used to help control the perceived loudness of TV adverts... so I expect some element of this monitoring of microdynamics will be introduced into the integrated measurement algorithm in time when it has been evaluated and developed further.

After all, the BS1770 loudness measurement is -- and can can only ever be -- an approximation of the hugely complicated sense of human perception. And human perception also varies considerably between different humans anyway!

I can't remember whether YT actually uses the BS.1770 algorithm -- some streaming services have their own variation on the theme. nevertheless, BS.1770 is easily the best approximation to perceived loudness that we've ever had -- although I don't think anyone would claim it's a perfect solution yet under all circumstances. That's why we're now at BS1770.4 -- the fourth iteration of the algorithm! And it is still being refined with the benefit of experiences exactly like the apparent divergence you've identified here.

Do you agree with me that it is nonsense that the streaming services use loudness normalisation...

No, I don't. And for several reasons.

The first reason is that regardless of whether it actually works perfectly for matching perceived loudness, or not, what it does do -- unquestionably -- is very strongly discourage peak-normalisation and the whole 'loudness war' mantra that truly was a 'nonsense' and hugely destructive to so much recorded music.

The second reason is that it does -- unquestionably -- provide a reasonably consistent balance between the perceived volume of widely disparate tracks. It's not perfect under all situation, granted, but it is a lot better than no normalisation at all. And that matters a lot because there is no-one controlling the output of all these streaming services (and increasingly, broadcast services too). It's all completely automated. So the only person to hear the output is the listener, and a lot of people made a lot of complaints about having to adjust the listening volume at every track junction. They don't do that anymore... so it's a win for the streaming companies!

The days of a chap wearing a tie and tweed jacket sitting in a 'continuity suite' listening to and manually balancing everything being broadcast in real time are long gone. So loudness normalisation is the next best thing. Not as good, true, but far better than nothing.

Yes, of course there are ways of defeating it for those selfish and arrogant enough to want to try. And no, it's not a perfect solution under all circumstances, and it can get caught out occasionally, and even get in the way occasionally...

But personally, I think it's more of a good thing than a bad thing...

YMMV
User avatar
Hugh Robjohns
Moderator
Posts: 30652
Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Two songs with exactly the same LUFS but one still sounds louder?!

Postby Dennis J Wilkins » Fri Nov 13, 2020 7:42 pm

By coincidence I have been working with sound/loudness perception while testing some multiband compressor software, and Matt and Hugh have hit the nail squarely — our ears/brain tend to perceive more clarity and excitement when there is more micro-dynamics even when the LUFS measurement reads the same. I’ve used a multiband compressor for years when the arrangement of a mix hasn’t already provided such “clarity”. This can add sonic variation by pulling up and pushing down different frequency elements throughout a song that is not obvious as manipulation, but is perceived as improved clarity, and it seems, as increased loudness.

I have a test project set up right now that enables me to switch between the original audio track (full stereo mix) and the same track processed by one of three multiband compressors. I can adjust the LUFS level, as measured by a Klangfreund Multimeter, to be the same over a musical passage across all tracks, and switch instantly between no compression and one of the processors. With five to six bands of compression, each operating within its own frequency band, the results are almost always significantly improved clarity and usually a “louder” sounding mix. Oddly, the actual compression ratios and thresholds are not critical to creating an improvement, although the actual sonic results vary depending on settings. BTW, I always use such compression in a parallel mode so as not to “squash” the dynamics.

I agree with Matt and Hugh, that loudness normalisation works pretty well — and with the cited examples, the better mix sounds better.
User avatar
Dennis J Wilkins
Poster
Posts: 25
Joined: Fri Aug 05, 2011 12:00 am
Location: Colorado USA
Dennis - Colorado USA