You Need Loudness Correction

[wadude]

The main reason I listen through speakers is that the sound changes based on many variables, not just loud and soft. It seems closer to the real thing.

 

[dasdoing]

There's not much loss with lower volume in the higher frequency part of the contours, so perhaps one only needs to use their bass control?

we actually have to look at the difference:

 

[Daverz]

we actually have to look at the difference:

View attachment 461619

Thanks. Where is that from? I'm not finding it enlightening.

 

[Hayabusa]

Thanks. Where is that from? I'm not finding it enlightening.

Indeed, not very clear what deltas, they are also off it seems.

These are the deltas measured from Audyssey wrt to the 85dB reference level for different volume offsets:

 

[Daverz]

Indeed, not very clear what deltas, they are also off it seems.

These are the deltas measured from Audyssey wrt to the 85dB reference level for different volume offsets:

View attachment 461648

Thanks, that's more helpful. Camilladsp has a loudness filter with a magnitude response that looks like this example:

But it does allow you to set separate low and high boosts, they do not have to be the same:

filters:
  loudness:
    type: Loudness
    parameters:
      fader: Main (*)
      reference_level: -5.0
      high_boost: 10.0 (*)
      low_boost: 4.0 (*)
      attenuate_mid: false (*)

More here: https://github.com/HEnquist/camilladsp?tab=readme-ov-file#loudness

Camilladsp also has a compressor, but that has a lot of parameters to adjust.

GitHub - HEnquist/camilladsp: A flexible cross-platform IIR and FIR engine for crossovers, room correction etc.

A flexible cross-platform IIR and FIR engine for crossovers, room correction etc. - HEnquist/camilladsp

github.com

 

[Hayabusa]

Thanks, that's more helpful. Camilladsp has a loudness filter with a magnitude response that looks like this example:

But it does allow you to set separate low and high boosts, they do not have to be the same:

filters:
  loudness:
    type: Loudness
    parameters:
      fader: Main (*)
      reference_level: -5.0
      high_boost: 10.0 (*)
      low_boost: 4.0 (*)
      attenuate_mid: false (*)

More here: https://github.com/HEnquist/camilladsp?tab=readme-ov-file#loudness

Camilladsp also has a compressor, but that has a lot of parameters to adjust.

GitHub - HEnquist/camilladsp: A flexible cross-platform IIR and FIR engine for crossovers, room correction etc.

A flexible cross-platform IIR and FIR engine for crossovers, room correction etc. - HEnquist/camilladsp

github.com

This example makes quite big corrections not really inline with equal loudness deltas I would say....
Room for improvement there.
The nice thing about Audyssey's Dynamic EQ is is has NO compressor, its purely only changes by the volume setting and as far as I can see a reasonable approximation of equal loudness curve deltas.

 

[dasdoing]

we actually have to look at the difference:

View attachment 461619

ok guys, to clarify. These are the compensation curves, so inverted deltas.
They are from: https://github.com/dpapavas/alsaloudness,
which also has a very good explanation of the concept.

 

[dasdoing]

I once created an audio-file in another topic that shows the effect. it's not on my Google Drive anymore, but if there is interest I can create it again

 

[HenrikEnquist]

This example makes quite big corrections not really inline with equal loudness deltas I would say....
Room for improvement there.

That example is set up for explaining how the loudness filter works. The corrections are too large on purpose, because that way it gets easier to see what happens in the figure.

 

[TheZebraKilledDarwin]

It's only potentially 'correct' for a limited number of movies made in the last decade or two and there are still way too many unknown variables for it to make any meaningful difference. And what if I simply prefer less bass than the producer did?

It's even dangerous to the hearing.
I have an Audyssey AVR and a good sounding HT setup, but at 0 dB MV (the calibration reference level Audyssey uses) I probably would become deaf.

What is even more dangerous, are many of the movie soundtracks themselfes.
There's often no quality control for audio of the consumer releases. A cheap intern may be responsible for the final audio preparation of the consumer release. And instead of adequate loudness parameters, just a limiter(!) might be smashed on the center channel. With the effect that the loudness of loud scenes even increases...

Some time ago I watched an action movie and there was a scene in a full and loud stadium. The scene lasted probably 10 minutes and the stadium noise reached 95 dB(A)! in my room - I set the volume on the AVR to have the dialogue around 55 dB(A). That's a LR (loudness range) of A-weighted 40 dB! A-weighted! That's without bass! For consumers in small rooms. Absolutely sick. Criminal.
Most people will not quickly enough react, that something is too loud. Especially audio ethusiasts often tend to explain problems away ("can't be too loud, my setup is calibrated and this is a Atmos diamond platinum master of the director's cut").

So it is VERY DANGEROUS to trust ANY of these "norms" or any technical claims. If you have one medium, where you do not know, what really was done to the soundtrack, and that is usually the case if you are not making the track yourself but consuming a finished product, you must be careful.
Trust your ears, not the labels and product claims and forget any "calibrated" levels.
The calibration is important and essnetial on the production side, where every segment of the chain is known.
As soon as you are consuming something, forget any "professional calibration" of your setup and only trust your ears.

 

[TheZebraKilledDarwin]

Says who? That's a very dubious claim that movies are 'meant' to be listened to 15 dB louder than music.

For me, any recording (whether movie or music or anything else) is essentially a black box of which we can know nothing with regard to the loudness, tonal balance, spatial chracteristics etc. of the recording situation. Likewise, it's utterly pointless to try and second guess the 'intent' of the artist/producer. All we have is the recording as it stands and the reality of our listening environment which we can then tweak so the subjective experience is as pleasing as possible for the widest variety of material. By all means use automated loudness correction if you feel it adds value to your experience, but don't imagine that it's in any way more accurate or 'correct' than manually adjusting tone controls to taste ...

While its true that modern music is much louder than movies, and therefore the volume level needs to be reduced, I think the oversimplification Audyssey uses to sell their system, is doing more harm than good. In facts I find it dangerous, because of the immense loudness differences among media:

Audyssey calibrates a single speaker to 75 dB SPL(C) with a pink noise signal level of -30 dBFS. That makes 105 dB SPL(C) per speaker if a signal level at full scale would be present.

Now modern music mixes easily have -7 LUFS. That's a PERCEIVED LOUDNESS only 7 dB lower than the theoretical signal peak level @0 dBFS of 105 dB SPL.

With modern music at -7 LUFS -> 105 dB - 7 dB = 98 dB of PERCEIVED LOUDNESS @0 dB Master Volume on the AVR per speaker.

Turning that MV only 15 dB down for listening to modern music, as Audyssey's rule of thumb suggests: 98 dB -15 dB = 83 dB PERCEIVED LOUDNESS! Per speaker.
Because usually two speakers are used for music, that would result in 86 dB PERCEIVED LOUDNESS. For music listening in small rooms. That's very loud in small rooms with good acoustics. But most rooms are untreated and decay times are above 500 ms.

 

[IPunchCholla]

So if one is in the Mac world, is there an easy way to test out if using loudness compensation is for me? I’m running Sequoia and have Sound Source.

 

[Mike123]

I listen to music quietly or very quietly. A comfortable quality level is 70-75 dB SPL. It can be louder, but not for long: 1-2 tracks. I feel uncomfortable for a more long time. Daytime background music is 10-20 dB quieter. Nighttime background music is 30-40 dB quieter. The brain probably adapts perfectly, but I have never had the desire to use loudness compensation.

 

[dasdoing]

So if one is in the Mac world, is there an easy way to test out if using loudness compensation is for me? I’m running Sequoia and have Sound Source.

give me a random song you love and know well and I create a small presentation. should have bass though

 

[IPunchCholla]

give me a random song you love and know well and I create a small presentation. should have bass though

How about A Hairdryer by The Smile?

 

[sceptical1]

Indeed, not very clear what deltas, they are also off it seems.

These are the deltas measured from Audyssey wrt to the 85dB reference level for different volume offsets:

View attachment 461648

This is what I hear. Not the fletcher munson or equal loudness curves, both highly exaggerated in my opinion

 

[Hayabusa]

This is what I hear. Not the fletcher munson or equal loudness curves, both highly exaggerated in my opinion

These deltas are still an (almost) direct conversion from the equal loudness curves.

 

[ZolaIII]

This is from work that annalise equal loudness trough time in 100+ scientific work's since 1983 up to now.

All the works where done on small to moderate number of participants and done on headphones/speakers or both. All together as results line in most part represent only scientific study in audio science with statistical significant and representive sample!
In order to simplify and make this easier to understand for most people I will focus on ISO 226 implementations of it. More specifically most widely used today ISO 226 2003 implementation and ISO 226 2012 and later implementations. Basic difference is additional boost in sub bass area in later implementations. And all do it's true and correct unfortunately we still lack technical capabilities to compile it entirely like that there as there is no subwoofer driver capable of doing it in the existence. However it's possible to use room fundamental to the length of min 8.5 m (preferably 9~9.5 m) in controled manner as compensation. So for the most folks ISO 226 2003 without that will still be perfectly fine.
Now to adresa different loudness levels in different material types and different DR boundaries. Best way to get to the bottom of this is law enforcement! And probably only one. The EBU R128 is EU broadcast association standard to deal with it with, with normalisation boundary of - 23 LUFS (dB) with acceptable margin of error of ±1 dB. With that error margin it's suitable even for TXH cinema 24 DR margins. In EU it's enforced now so all broadcast materials should have to be compliant with it (which still really is not the case). It's available for many players/DAW's in various plugin forms and mostly as free one so it's not a problem to make your own private collection to go compliant. It's also not a problem to compensate for it's average loudness reduction (about - 11 dB) not introducing any noise to the signal by simple using FP domain math for it (as SINAD is far above any sain hearing threshold there) so that you don't need insane preamp stage to do it not losing on your pressies SPL output or needing insane amp/power amp.
I hope this in short address and explain all of your previous asked questions throughout this thread as I whose constantly repeating this all over the ASR for years now.
There are other complications in implementing ELC especially when amount of compensation (boost) becomes significant (10~12 dB or more) for program listening levels of 60~63 dB (normal come speach) or even lower ones. It mainly comes to the knew of ELC for low/sub bass part is at 105 Hz and only way ensuring harmonics from it's boost to stay under control is putting crossovers above it to 120 Hz. This leads to 2.2 setup and has nothing to do with something as silly as stereo bass! For average programme SPL most of us like to listen at front 70~76 dB where boost is less severe you can pass with single bass chenel and lot lower crossover (that sub/sub's basically do their part on low/sub bass).
Hopefully this explains it to you.

 

[jkim]

This study:

https://www.smc2019.uma.es/articles/S2/S2_01_SMC2019_paper.pdf

suggests loudness compensation derived from the differences among ISO 226 equal loudness contours.

Interestingly, they chose not to compensate for treble as the differences are very small. But as we know traditional loudness control involves a boost in treble.

Did they miss anything about human loudness perception? Perhaps something that equal loudness contours cannot fully capture?

Perhaps @solderdude may also chime in here as he took a similar approach as mentioned in this thread.

 

[jkim]

we actually have to look at the difference:

View attachment 461619

How did you derive these curves? According to the differences (deltas) in equal loudness contours, the compensation should be mainly on the bass response---also see the study I posted above as well as @solderdude's compensation curves (also Audyssey 's delta curves posted by @Hayabusa).