Here's something I've been working on with MeterPlugs recently.

Find out if your music will be turned down by YouTube, Spotify, TIDAL, Pandora and Apple Sound Check - discover your music's Loudness Penalty score, for free.

There's a demo video here:

http://productionadvice.co.uk/loudness-penalty/

Or, you can go straight to the site itself

http://www.LoudnessPenalty.com

Any questions let me know !

Another really useful tool, Ian. and FREE too! Nice one.

H

Ok, I'll ask a numpty question here : how much does it matter?

Does the reduction in volume impact the overall listening balance much?

I know that over compressed tracks might result in a lower perceived volume when the level is reduced but I thought that with the advent of LUFS based mixing targets, we should be reasonably dynamic and shouldn't suffer much in that regard.

All good tools are welcome (doubly so for free ones) so is this more for those who haven't got the message?

(quite possible I've missed a meeting somewhere ... )

Dave B wrote:Ok, I'll ask a numpty question here : how much does it matter?

Depends on what your source mix is like. If it's a crushed-to-death master, it will be 'punished' a lot (ie, turned down and come out sounding feeble and quiet). Conversely, If it's a well-balanced, dynamic track mixed to hit the appropriate target level, it won't be 'punished' at all.

And that's what this online measuring tool indicates.

I thought that with the advent of LUFS based mixing targets, we should be reasonably dynamic and shouldn't suffer much in that regard.

Absolutely true... but not everyone is as well educated and enlightened as the SOS readership!

...so is this more for those who haven't got the message?

Basically -- yes! As Ian says in his supporting video, it's a very useful tool to help persuade clients or A&R agents who still don't get the whole ethos of loudness normalisation and keep asking ' can you make it louder?'.

H

Nice one Ian!

Fab tool.

Thanks chaps !

Hugh is right - the degree to which the level reduction is a sonic penalty as well as numerical one depends on the material. In my experience, LP below -2 on YouTube is the point where things can start to suffer - compare the examples in this playlist, for example:

http://bit.ly/lp-in-action

"Broken Beat & Scarred" has an LP of -7.9 (!) and has no-where near the impact of "Hardwired", which scores LP -2.4, for example.

On the topic of LUFS "targets", I've found that aiming for a particular integrated loudness (-14 LUFS, say) isn't the best strategy. More info here:

http://productionadvice.co.uk/how-loud/

Ian

Many thanks Ian!

Having digested your 'how loud' article as well as trying a couple of my recent tracks on this most excellent web page, I can see that my approach of aiming for around -16 LUFS (with True Peaks no higher than -1) is a bit conservative, giving my tracks a -4 to -5 (i.e. I could raise their integrated loudness by at least 4dB and not be penalised by streaming sites).

In future I'll work towards your stated aim to "Master no louder than -9 LUFS short-term at the loudest moments - currently these same tracks indicate Short-term loudness of about -13 LUFS, which ties in perfectly, meaning I can raise the integrated loudness by about 4dB. Spot on!

Thanks again


Martin

Martin Walker wrote:Having digested your 'how loud' article as well as trying a couple of my recent tracks on this most excellent web page, I can see that my approach of aiming for around -16 LUFS (with True Peaks no higher than -1) is a bit conservative, giving my tracks a -4 to -5 (i.e. I could raise their integrated loudness by at least 4dB and not be penalised by streaming sites).

In future I'll work towards your stated aim to "Master no louder than -9 LUFS short-term at the loudest moments - currently these same tracks indicate Short-term loudness of about -13 LUFS, which ties in perfectly, meaning I can raise the integrated loudness by about 4dB. Spot on!

Thanks again

My pleasure - sounds like a plan

So the only thing I'd stress is that people should actually preview their music against other similar material on YouTube (or wherever) to make sure it doesn't already sound loud enough musically, even though it's not being turned up. For example ambient or gentle acoustic material may not need to sound as loud as everything else.

But with a decent limiter, increasing the level by a few dB will almost certainly be possible with very transparent results.

Having said all that, the AES guidelines for streaming are a maximum of -16 LUFS, so if the services standardise at some future point, we may change our minds about what's best again...!

Ian

I just finished a song that I submitted to an FM radio station for air play. When I put the master into this site it reads this:

--
YouTube

-1.3
Spotify

0.0
TIDAL

-0.8
Pandora

-2
iTunes*

Can you help me understand what FM radio is similar to? Would it be possible to have a loudness penalty for FM radio? Does this even apply?

I really hate the sound of squashed bus compression and tried to make the song sound dynamic and lively just using my ear. But I have not heard what this song sounds like over FM radio yet. I am always afraid that it will sound whimpy compared to the previous song, but will have to leave that up to the station to get right.

DC-Choppah wrote:Can you help me understand what FM radio is similar to? Would it be possible to have a loudness penalty for FM radio? Does this even apply?

I really hate the sound of squashed bus compression and tried to make the song sound dynamic and lively just using my ear. But I have not heard what this song sounds like over FM radio yet. I am always afraid that it will sound whimpy compared to the previous song, but will have to leave that up to the station to get right.

FM processing is a whole other kettle of fish :/

Online streamers just use a simple gain change, whereas most FM stations use aggressive multiband processing on everything, which can significantly change the sound, regardless of the mastering decisions we make.

The good news is that less aggressive processing almost always sounds better, on FM. You can hear a real-world example here:

http://productionadvice.co.uk/loudness- ... the-radio/

If your song is an aggressive genre, or intended to sound loud from start to finish, you might want to experiment with pushing it a little harder and see how it sounds in comparison to other similar songs, to make sure you have the right "density". But if it has varied dynamics (verse/chorus etc) and is meant to have some light & shade, you're almost certainly already in great shape.

Great examples on your site man!

What I hear on your examples of the three levels of mastering is that the FM broadcasts squash the drums more and more and they lose their punch and natural sound the more the original mix wash squashed. Really takes out the emotion of the drums. And the natural arrangement balance between the instruments gets lost and normalized. All the instruments sound like they are more an more at the same level. Much less musical.

There was this important tidbit I saw on your site too: 'The broadcast processing removes all the internal dynamics of the song – the beginning is just as loud as the end'

So attempting to level the verses and choruses with macro level changes does not read on FM broadcast.

But for streaming, the macro leveling will read, and you can add a little drama with level changes over the track since they only apply a single gain adjustment to the whole track.

Did I get this right? Just following my ears.

DC-Choppah wrote:Great examples on your site man!

What I hear on your examples of the three levels of mastering is that the FM broadcasts squash the drums more and more and they lose their punch and natural sound the more the original mix wash squashed. Really takes out the emotion of the drums. And the natural arrangement balance between the instruments gets lost and normalized. All the instruments sound like they are more an more at the same level. Much less musical.

There was this important tidbit I saw on your site too: 'The broadcast processing removes all the internal dynamics of the song – the beginning is just as loud as the end'

So attempting to level the verses and choruses with macro level changes does not read on FM broadcast.

But for streaming, the macro leveling will read, and you can add a little drama with level changes over the track since they only apply a single gain adjustment to the whole track.

Did I get this right? Just following my ears.

Glad it was helpful - and yes, you got all that right. The only thing I'd add is that there is significantly more distortion in the FM processed version which was loudest "going in". Lots of people think they need to squash stuff hard to prevent this happening, but it actually just sounds worse.

And another thing to point out is that even the loudest example in my post isn't super-loud, by modern standards. DR6 and DR 4 masters are all over the place now, and they sound even worse after FM processing

And yes, macro-dynamic contrast is maintained online, but not through FM processing.

Last thing to add - not all FM stations use this kind of processing, some use level balancing like the online streaming services, and a few use nothing at all.

So the bottom line is that you can't 'win', so as always my advice is to aim for the sweet spot - loud enough but not too loud, as described in my "how loud" link above.

Hope that helps !

Fabulous work Ian! A great resource, and I'll certainly be shouting about it to my lists.

One slightly tangential question, if I may: to what extent do we know the technical mechanics of the loudness-detection algorithms used by these services? For example, I've been reading the ITU BS1773 spec, and wondered whether the streaming services use something similar to the K-filter (HPF and high-shelf) pre-emphasis mentioned there?

May I raise a small, but important, technical point, Mike.

Mike Senior wrote:I've been reading the ITU BS1773 spec...

There isn't an ITU BS1773 spec! Although ITU BO1773 does exist and refers to criterion to assess the impact of interference to broadcasting-satellite services...

I think you're actually referring to the ITU-R BS1770-4 specification which relates to loudness normalisation.

ITU-R means it's an ITU Radiocommunications section recommendation. BS means it's in the Broadcast Sound section. 1770 is the core specification document. And -4 means it's currently on its fourth published revision.

So, ITU-R BS1770-4, please!

... and wondered whether the streaming services use something similar to the K-filter (HPF and high-shelf) pre-emphasis mentioned there?

I suspect they must be: the P-Loud group that developed the system found that that pre-emphasis curve is critical in the achieving good correlation between the metered value and the actual perceived loudness.

However, even quite small differences in the filter shape will change the metered value by a few decibels, and I suspect that is the reason why the different streaming services appear to have slightly different target loudness values when assessed against the BS1770 spec.

H

Hugh Robjohns wrote:So, ITU-R BS1770-4, please!

Schooled again... Slip of the keyboard on my part, although I'd still have missed out the 'R'... Thanks for the clarification, Hugh!

I suspect they must be: the P-Loud group that developed the system found that that pre-emphasis curve is critical in the achieving good correlation between the metered value and the actual perceived loudness.

However, even quite small differences in the filter shape will change the metered value by a few decibels, and I suspect that is the reason why the different streaming services appear to have slightly different target loudness values when assessed against the BS1770 spec.

Interesting. Not my specialist subject, of course (as is already apparent!), but I do like to keep abreast of some of the technological issues, so thanks again for that extra info!

Hi Mike,

Thanks for the positive feedback, and good question ! The answer is a little messy:

TIDAL use loudness units (LUFS), and we know their method - they match the loudest song on each album, and scale everything accordingly, even in shuffle play. This is the best method, in my opinion - it retains the musical relationship between songs, which is what 80% of listeners prefer. So an acoustic ballad won't sound as loud as a death metal song, for example.

Spotify use ReplayGain, which is the same idea as LUFS but less effective, in my experience. They use track normalisation (per song) in shuffle mode, and album normalisation (retaining musical balances from song to song) for albums and playlists.

There's no official word from YouTube or Pandora, but our numbers are very close

Apple use their own Sound Check algo, which is a law unto itself - it's the least accurate of our reported values - but we're working on it ! They also have an "Album mode", although it can be a little flaky at times.

Summary: LUFS can give you a helpful ballpark figure, but it's not sufficiently accurate except in the case of TIDAL, which is why we created the Loudness Penalty site. Overall most sites normalise to roughly -14 LUFS, but I don't recommend that as a target - see the How Loud post I mentioned earlier.

Ultimately it's a confusing business, and we can hope that eventually they'll all move towards TIDAL's solution - but that's a way off, for now.

The good news is there's no pressure to push the loudness really hard online, we can do what sounds best for the music (imagine !)

Ian

Ian Shepherd wrote:The good news is there's no pressure to push the loudness really hard online, we can do what sounds best for the music (imagine !)

Whisper it, Ian, whisper it...