Device for Equal Loudness/Fletcher Munson Curve? Do Any Speakers Adapt to This?

[Rock Rabbit]

the thing is, there is no perfect attenuation. nobody said the proposed method is perfect. what is beeing said is that it is much, much better then simply attenuating all frequencies equaly.
at this point it realy makes more sense to hear the result, then to argue over the theory.
if you have any method that does it even better, I will be happy to use it

Some people propose simply lifting the bass with shelving EQ fatally spoiling harmonics content. The most flawed method seems to be popular.
Some gear comes with "continuous variable loudness"...and not very popular.
There's no perfect solution, but there are fatally flawed ones that many use!.
The trap always start with "true to the source". My very personal "solution": with near field monitors no EQ (very obvious), for typical room LS, the room gain and modes is enough mess for me (and LS low freq at the limit of excursion as any LS) then there's no room for EQ. Headphones is another thing and there's more freedom if excursion and THD is under control...it seems perfect for magnetoplanars, I use EQ for resonances and dips and very small shelving EQ in bass range. For reference headphones never!
In many situations the "solution" is worst than the problem...damped small room with good small speakers (due to room gain) and near field monitors for critical parts that's all, less is more. Your ears + brain DSP takes the room out of the equation without eq (famous Dr. Toole study)
https://www.audioholics.com/room-acoustics/room-reflections-human-adaptation

 

[ahofer]

I'll just plug the RME version built into the ADI-2 DAC (again). You can set the floor/ceiling level and the amount of change in treble and bass with increases in signal voltage. You can also set one of the remote buttons to defeat/engage it. IMO, this is how it should be implemented to avoid many problems mentioned above.

The ADI-2 DAC offers Loudness for the analog stereo outputs, and probably is the first time that Loudness works as intended. The user can decide how much maximum gain in Bass and Treble should occur at lower volume settings. The user also sets the Low Vol Ref- erence, where maximum gain is achieved. After extensive tests a 20 dB range has been defined as range for maximum gain to no gain while increasing volume. That seemed to be the perfect definition of the range that needs to be addressed by Loudness.

Here is an example on how it works: the user’s typical lowest level listening volume is at -35 dB at the unit. This value is now set by the user as Low Vol Ref in the Loudness menu. Then Bass and Treble Gain can be set between 0 and +10 dB. Default is +7 dB for both. Increasing the volume by turning the Volume knob causes the gain in Bass and Treble to be lowered smoothly over a range of 20 dB. So when Volume is set to -15 dB, the music is not only quite loud, but Loudness’ Bass and Treble are then at 0 dB gain. See chapter 31.8 for graphs.

No matter how sensitive the connected phones or speakers are, no matter how much increase in Bass and Treble are desired – with the ADI-2 DAC one can finally adjust it to meet the per- sonal hearing and taste. Loudness finally works as it should have worked from the start - anoth- er unique feature in the ADI-2 DAC.

https://www.rme-audio.de/downloads/adi2dac_e.pdf

 

[ernestcarl]

Your ears + brain DSP takes the room out of the equation

Yeah, up to a point.

What are you proposing for speakers exactly? No external DSP other than the one already in-between the ears and brain?

 

[Rock Rabbit]

Yeah, up to a point.

What are you proposing for speakers exactly? No external DSP other than the one already in-between the ears and brain?

Or maybe an external DSP that don't change the relative harmonic content of music (the bare minimum) some coarse eq movements and this minimum disappear! An example from YouTube "small bell inharmonic spectrum 1 kHz and up to 20 kHz

What loudness curve is valid so the relative inharmonics are preserved in amplitude?
All that I say is that simple EQing is the wrong answer and an average loudness curve could start a storm like the Harman curve for headphones (?) ...but a good DSP curve (maybe with SPL feedback, several mics in the room) is in the correct path
BTW: it seems that Apple headphones EQ is with mic feedback. So maybe the idea has some marketing potential ;-)

 

[Aerith Gainsborough]

What loudness curve is valid so the relative inharmonics are preserved in amplitude?

They don't need to be. That is the entire point.
You are measuring the signal before the actual sensor which displays a non linear behavior.

Expecting the ratios to stay the same is absurd.

 

[Rock Rabbit]

They don't need to be. That is the entire point.
You are measuring the signal before the actual sensor which displays a non linear behavior.

Expecting the ratios to stay the same is absurd.

The ratios define the timbre on any instrument so you can identify them. Loudness curve is nonlinear, then add amp (linear) then loudspeaker (non linear in full range) then room modes (very non linear high Q)....and then your ears threshold curve makes all that flat? Or at least nearly flat to at least preserve instrument timbre? At least that's the idea of flat response in the digital analog electronic chain. The same idea of "flat" loudspeakers in anechoic test, flat near field monitors to make music production...the only nonlinear sensors are our ears and room interactions.
If that last chain link must be compensated to "true sound" it needs to follow the same basic principle of the digital/analog chain.
Loudspeakers are very non linear, then they are restricted to work band limited with xover and compensated for efficiency and directivity....that's what we do with non linear transducers

 

[Aerith Gainsborough]

I was talking about your EAR.
That is a non linear sensor. So, to expect a loudness compensation to maintain the relative amplitude of the harmonics means you fundamentally misunderstand the point of the compensation in the first place.

Yes, Room, Speakers etc alter the sound already. That can be remedied so a certain degree via DSP. Especially room modes can be controlled that way or at least brought down to +-5dB.

What is left is the non linearity introduced by your ear. You need to differentiate the signal before the sensor (a.k.a.: what you loudspeaker emits) and the signal after the sensor (the electrical signal that goes to your brain). The goal of the compensation is to maintain the relative harmonics in the latter signal by altering the relative contents of the former.

Naturally, that cannot be exactly measured, since we cannot stick a probe into our brains (yet, give Amir a few decades). So there is always a point at which the listener has to fine tune the compensation until it sounds right.

 

[Rock Rabbit]

There's no easy solution because the DSP or VST solution needs to know (pre process) the real loudness of the track, second problem it needs to know the real SPL vs volume setting of the actual loudspeaker. For audio production the near field monitors must be calibrated and set at ONE calibrated volume. For consumer no difference but the software must adapt curves to different volume (loudness settings). Read:
https://www.gearslutz.com/board/mus...est-plugin-defeating-loudness-perception.html
https://www.audiosciencereview.com/...dness-contour-loudness-dsp-for-playback.5955/

P.S.: I really hope that people in this thread could really understand that the solution is not the use of arbitrary compensation "by ear"...that's not the idea behind equal loudness contour

 

[stevenswall]

There's no easy solution because the DSP or VST solution needs to know (pre process) the real loudness of the track, second problem it needs to know the real SPL vs volume setting of the actual loudspeaker. For audio production the near field monitors must be calibrated and set at ONE calibrated volume. For consumer no difference but the software must adapt curves to different volume (loudness settings). Read:
https://www.gearslutz.com/board/mus...est-plugin-defeating-loudness-perception.html
https://www.audiosciencereview.com/...dness-contour-loudness-dsp-for-playback.5955/

P.S.: I really hope that people in this thread could really understand that the solution is not the use of arbitrary compensation "by ear"...that's not the idea behind equal loudness contour

Simply have the device analyze the loudness in real time which takes effectively zero processing power. Have the user play music as loud as they are going to and set a curve. Then do the same for as soft as it's going to play and set that curve. The software/DSP device can then adjust to intermediary levels as it is used, taking a fraction of a second to analyze the audio files it is fed, or doing something like a trailing 5 second average or real time adjustment.

 

[Rock Rabbit]

Simply have the device analyze the loudness in real time which takes effectively zero processing power. Have the user play music as loud as they are going to and set a curve. Then do the same for as soft as it's going to play and set that curve. The software/DSP device can then adjust to intermediary levels as it is used, taking a fraction of a second to analyze the audio files it is fed, or doing something like a trailing 5 second average or real time adjustment.

If you compute loudness in real time, how you know the loudness ahead in time? What if the song ends loud?
Effectively for streaming or broadcasting you can set a loudness target in real time, but destroying the real "canned" loudness of the music...no good for "real" music (loudness must be calculated for every track ahead, not real time but almost).
How loud it's gonna be played? Or there are calibrated speakers (or at least 1 point calibration with mic, or a table with volume vs effective SPL output) or an integrated system with calibrated mic feedback...certainly it can be done (with loudness pre process this stage is real time).
At least my old i5 takes forever calculating BS1770 loudness....but for a dedicated DSP it must be some seconds per track

 

[stevenswall]

If you compute loudness in real time, how you know the loudness ahead in time? What if the song ends loud?
Effectively for streaming or broadcasting you can set a loudness target in real time, but destroying the real "canned" loudness of the music...no good for "real" music (loudness must be calculated for every track ahead, not real time but almost).
How loud it's gonna be played? Or there are calibrated speakers (or at least 1 point calibration with mic, or a table with volume vs effective SPL output) or an integrated system with calibrated mic feedback...certainly it can be done (with loudness pre process this stage is real time).
At least my old i5 takes forever calculating BS1770 loudness....but for a dedicated DSP it must be some seconds per track

For music real time, 500ms, and 1 second don't matter... You don't need to know the future, just process stuff and that's that. If it ends loud, the user can decide if they want keep the same curve or have it adapt more quickly at different points in the song.

Anyhow, it sounds like a great solution hasn't been made yet, even though all the technology needed to do so is available. Unfortunate. Also unfortunate an iPhone can't make a 3D model of a face and calculate a unique HRTF for the user and apply time delayed and EQed crossfeed to headphones based on the model and pair of headphones being used.

 

[Aerith Gainsborough]

Simply have the device analyze the loudness in real time which takes effectively zero processing power.

That wouldn't work and that is also not the point of such a compensation.

If you have a quiet passage in a track, that passage is intentionally quiet (including the altered spectral balance perceived by your ear).
One should not try to compensate for that.

Same goes for different tracks within the same album. Usually, there are artistic reasons for volume differences.

You really only need the peak levels in relation to the calibration point. EG: If the track peaks @ 85dB on your system in the current setting, no calibration should be applied. If it peaks at 60dB because you turned the system down by -25dB, you compensate for these -25dB and that's that.

You could do the same with the perceived average loudness or "LUFS" metric instead of peaks, whatever you personally prefer. In a home music setting with neighbors, most people adjust the volume to the actual peak values though, so both approaches may have merit.

It's not exactly rocket science, to be honest. There is no need to overcomplicate things and try to compensate to an absurd degree.
Sometimes, less is more.

 

[dasdoing]

note that loudness normalization and equal loudness correction are two distinct things. it is true that the second does not make too much sense without the first. the first problem is generic though, since it is a problem for anybody "claiming" to listen at reference level only, too. the fact is we first have to "create" a fixed reference level for the music we play. the best tool for this atm is bringing everything to meassure the same integrated LUFS level.

 

[pkane]

Are there any VST plugins that do a good job of equal loudness compensation using F-M curves? I'm almost tempted to build one myself, just to play with it

 

[Rock Rabbit]

Are there any VST plugins that do a good job of equal loudness compensation using F-M curves? I'm almost tempted to build one myself, just to play with it

https://vst4free.com/plugin/2239/
https://www.noisebud.se/?download=smile-v3
https://lsp-plug.in/?page=manuals&section=loud_comp_stereo
And some Foobar plugins

 

[Chromatischism]

If nobody has mentioned it (how does one search a particular thread?) the Buchardt A500 has a calibrated loudness feature you can turn on.

In the automotive space, BMW and others have speed dependent volume; I don't recall ever discussing if this was strictly volume, or if any loudness EQ was also involved

And the Buchardt amp also has it.

I just use Dynamic EQ because I'm cheap but it works great

 

[ahofer]

Going back to the RME, I don’t know why it wouldn’t work. I can’t find a technical explanation, but I believe the Loudness works in EQ. You set a reference level for maximum boost and it applies boost accordingly. My supposition is that it applies based on signal voltage level vs. reference. In that case it should work quite well, no?

 

[Aerith Gainsborough]

Well, now I have the RME (wee!), so a few quick observations:

Loudness works well, I measured my output (speakers and cans) and apply the boost below the volume setting that gives me 90dB (Z weighted) peaks. No clue if that's the right way to do it but it suits the listening levels I use most often.
It's much more comfortable to use than the JRiver implementation b/c it does not interfere with my VST setup.

There are caveats and pitfalls though:
Individual track volume is a PITA. Sure, replay gain helps but only to a certain extent. If Track A is so quiet compared to track B that you have to adjust the volume, you are basically screwed. You will get way more bass than intended on Track B because you toned it down.

Good example for me would be 2 tracks from the Space Engine OST:
A peaceful place: -15 LUFS (mostly low frequency content)
Eternal moment: -10.9 LUFS (much more high frequency content)

If I listen to the first one at an elevated level, I will have to turn the second one down. It's simply too loud. Setting Replay gain to "Track based" fixes this but that is not applicable to many albums, where you are supposed to have loud and quiet tracks.

The most extreme delta is -8.4 vs -20.9. Oof. Now this is probably because it is not so much a soundtrack (even though it is used in the Simulator) but rather a collection of individual tracks from different artists.

The easiest solution I found so far (short of normalizing the album), is to just manually tone down the volume in foobar instead of using the RME's volume. That way the loudness compensation doesn't kick in. I do wish there was an easier way to switch between replay gain modes in foobar. (never mind, found it!) Album is probably fine for most stuff but on some randomly cobbled together playlists "per track" works better.

 

[Weeb Labs]

I have actually designed almost exactly this device and am currently awaiting board fabrication. It is called the Voluminous and is essentially an overengineered loudness compensated active preamp with selectable curve slopes, I/O buffers to aid longer cable runs and bypass functions. It is based upon the NJW1194. Perhaps I'll open source this (and send one to Amir) once testing is complete.

 

[ahofer]

Well, now I have the RME (wee!), so a few quick observations:

Loudness works well, I measured my output (speakers and cans) and apply the boost below the volume setting that gives me 90dB (Z weighted) peaks. No clue if that's the right way to do it but it suits the listening levels I use most often.
It's much more comfortable to use than the JRiver implementation b/c it does not interfere with my VST setup.

There are caveats and pitfalls though:
Individual track volume is a PITA. Sure, replay gain helps but only to a certain extent. If Track A is so quiet compared to track B that you have to adjust the volume, you are basically screwed. You will get way more bass than intended on Track B because you toned it down.

Good example for me would be 2 tracks from the Space Engine OST:
A peaceful place: -15 LUFS (mostly low frequency content)
Eternal moment: -10.9 LUFS (much more high frequency content)

If I listen to the first one at an elevated level, I will have to turn the second one down. It's simply too loud. Setting Replay gain to "Track based" fixes this but that is not applicable to many albums, where you are supposed to have loud and quiet tracks.

The most extreme delta is -8.4 vs -20.9. Oof. Now this is probably because it is not so much a soundtrack (even though it is used in the Simulator) but rather a collection of individual tracks from different artists.

The easiest solution I found so far (short of normalizing the album), is to just manually tone down the volume in foobar instead of using the RME's volume. That way the loudness compensation doesn't kick in. I do wish there was an easier way to switch between replay gain modes in foobar. (never mind, found it!) Album is probably fine for most stuff but on some randomly cobbled together playlists "per track" works better.

Doesn't Roon's volume leveling feature help with this? I think similar features exist in other players.

https://help.roonlabs.com/portal/en/kb/articles/volume-leveling