Yup. Many funny issues I've pondered over the years... without a good conclusion. For the moment, I can't think of an analogy from visual perception."Loudness" compensation in playback equipment has always puzzled me since I think the ear/brain adjusts for this automatically. When I go to the Seattle Symphony does the orchestra need to "equalize" their sound when playing softly vice full force?Everything we hear in nature is not equalized depending upon absolute SPL.
Dave.
But in audiophile practice, I think it is fair to say that music at home sounds better when there's some boosting of the bass as the loudness goes down and the olde Fletcher-Munson curves (as realized by 4-pot "loundess" volume controls) fits nicely (assuming you've done the gain management needed).
Long, long ago, there was a false interpretation of the FM curves that suggested treble be boosted too. I am not sure how many people advocate a bit of treble tweaking too, regardless of the FM curves.
Ben
Last edited:
"Loudness" compensation in playback equipment has always puzzled me since I think the ear/brain adjusts for this automatically. When I go to the Seattle Symphony does the orchestra need to "equalize" their sound when playing softly vice full force?
The orchestra sounds very different at mp than they do at fff, though.
The idea of loudness correction is to get the listener to perceive the mp as mp and the fff as fff, even if the overall level is lower.
So, is this idea going anywhere?
The reason I ask is because I had the exact same idea just the other day and then I was quite happy to find that someone had already suggested it to the miniDSP team.
What I'd like is an automatic continuously variable loudness control (ACVLC!). I have no interest in using the miniDSP volume pot for this. Getting up off my butt to change a manual loudness control dial every time I change my main volume would render my pre-amp remote control useless.
If the miniDSP could simply apply a user-configurable amount of F-M curve based on the input signal level, it would be perfect. Set it and forget it. Great "loudness" response at all listening levels, without leaving your chair.
In the meantime, I simply configured EQ settings that sound excellent at my "average" listening levels. Bass is lacking at very low volume, but if the volume is that low then usually it is my wife listening and she would never notice. When I crank up the volume to much louder than normal, I just let the bass shake the house. Not ideal, but much better than tweaking the EQ every time I change the volume.
The reason I ask is because I had the exact same idea just the other day and then I was quite happy to find that someone had already suggested it to the miniDSP team.
What I'd like is an automatic continuously variable loudness control (ACVLC!). I have no interest in using the miniDSP volume pot for this. Getting up off my butt to change a manual loudness control dial every time I change my main volume would render my pre-amp remote control useless.
If the miniDSP could simply apply a user-configurable amount of F-M curve based on the input signal level, it would be perfect. Set it and forget it. Great "loudness" response at all listening levels, without leaving your chair.
In the meantime, I simply configured EQ settings that sound excellent at my "average" listening levels. Bass is lacking at very low volume, but if the volume is that low then usually it is my wife listening and she would never notice.
When I crank up the volume to much louder than normal, I just let the bass shake the house. Not ideal, but much better than tweaking the EQ every time I change the volume.
Last edited:
Granted, nobody I've ever heard from seems to understand how Fletcher-Munson works in real life, myself certainly included. But for sure, what you are talking about, if I actually understand what you are suggesting, seems inapplicable.
It is easy to utter the words,".... and when the music is soft, use the 80dB curve....". But it is quite another thing to have a transistor triggered for "soft" since even the loudest moments of music have silent instances.
It is possible to have a time-weighted averaging and many circuits do things like that. But that's not part of the Fletcher-Munson curve family and I have never seen that kind of "3D or "4D" data, even though it is in-theory possible.
To make a long story short, just as "dB flat" is a non-starter for tuning your music system, likely for loudness compensation from Fletcher-Munson curves too. In both cases, we can each choose "personal corrections" that make the music sound right to us. But that's personal taste. It is comforting to be able to "quote" curves to support our actions. But it is rare to be able to rest-your-case at that point.
If it is a matter of personal taste, no need to wait for somebody else to create a Fletcher-Munson-tuned device.
Ben
Last edited:
One year later, yet someone else investigating the idea (me). I think it is a really good one.
If you mean miniDSP should track the "live" input signal level - that is not what you want. It would be totally messing your dynamics (low level passages of the same track being played with different correction than loud ones).
What you want (as pointed out before in the thread) is a way to calibrate miniDSP to your SPL level at the listening position wrt. a given volume knob reference (obviously the one from the miniDSP being the straightforward way). Calibration only needs to be done once.
Yet better - but harder - the recording level should be also taken into account, but that's only "nice to have".
Like you, I would not use miniDSP volume control, which makes things significantly more difficult (how to tell miniDSP which position the volume control knob is).
"Loudness" compensation in playback equipment has always puzzled me since I think the ear/brain adjusts for this automatically. When I go to the Seattle Symphony does the orchestra need to "equalize" their sound when playing softly vice full force?
Dave.
Of course the orchestra does not have to equalize soft passages vs loud ones, because they play at the reference level.
Loudness compensation is not about equalizing low vs loud passages according to an absolute SPL, but about (statically) equalizing the sound when played back at a given lower relative level compared to the original so that the relative loudness of sounds at different frequencies is preserved.
It is inherently flawed - cannot be done 100% accurately, because this would require that:
1. the original (average or reference) loudness is known, so we can pick a reference equal loudness curve and do the correction towards the lower listening equal-loudness curve. Unfortunately, this information is not readily available.
2. all equal-loudness curves be equi-distant at all frequencies - this would ensure that the correction necessary when going from say 80phon to 60 phon is the same as going from 90 phon to 70 phon.
Without No 2, one cannot statically equalize one equal loudness level to another - it would be necassary to perform level-dependent EQ, a daunting task.
However, if we look at the newer ISO curves, No.2 seems to be remarkably well fulfilled: Fletcher?Munson curves - Wikipedia, the free encyclopedia
No.1 is a tough one and does not seem solvable automatically.
However, a DSP solution can still be significantly better than no correction at all.
EDIT: there's also a No.3 difficulty and that is the recording EQ.
Last edited:
I have abandoned the idea of using a miniDSP and I am instead going to build a passive variable loudness control circuit and integrate it into my pre-amp. The goal will be to integrate it in such a way that the volume buttons on my remote actually adjusts the loudness control. Volume will be preset manually to an appropriate level, just below "too loud".
I've been contemplating "loudness compensation" for 50 years and still don't understand it. For sure, not a "real time" thing ("live") which would be as aurally offensive as compression. But then what is the time course averaging of the operational level compared to a "reference level" or when an orchestra is playing a loud passage and you are playing it back not so loud?
In practice, I find loudness compensation (of the old, 4-gang volume control sort) to be pretty nice for late-night listening. But a distortion any other time even with the best of gain management. In as much as it is helpful only when music is just background wallpaper, it just doesn't matter getting it any too precise so long as it is about right at your usual low-level loudness. Like with a lot of otherwise valid acoustic theory, Fletcher-Munson curves (and speaker enclosure sims) aren't nearly as predictive of what sounds good as you might think.
Ben
In practice, I find loudness compensation (of the old, 4-gang volume control sort) to be pretty nice for late-night listening. But a distortion any other time even with the best of gain management. In as much as it is helpful only when music is just background wallpaper, it just doesn't matter getting it any too precise so long as it is about right at your usual low-level loudness. Like with a lot of otherwise valid acoustic theory, Fletcher-Munson curves (and speaker enclosure sims) aren't nearly as predictive of what sounds good as you might think.
Ben
You don't need time average, just the loudness ratio between live/playback at a reference frequency. Think about it this way:
- the original source (say a piano) plays a 1000Hz note at say loudness 60phon. According to the 60phon equal-loudness curve in the FM model, this requires 60dB SPL.
- same orchestra now plays a lower 100Hz note, equally loud: according to the 60 phon curve, at 100Hz it takes a higher SPL to generate the same loudness sensation, around 75 dB.
- now you are playing the recording of the above such that the 1000Hz note plays at 20dBSPL, 40dB lower that "live". The 1000Hz note is now 20 phon loud (corresponding to 20dB SPL at 1000Hz)
- this means that now the 100Hz note plays at 75-40 = 35dB SPL.
- however, in order to preserve the same 20 phon loudness of the 100Hz compared to 1000Hz you would need to follow the 20phon loudness curve and see that it would actually require ~45 dB for the same loudness. That is 10 dB higher than uncorrected !
Because the phon scale at ~1000 Hz roughly corresponds to the SPL scale, the correction you need when lowering the SPL from "live" levels to playback can be obtained by superimposing the 2 equal-loudness curves at 1000Hz and computing the SPL difference elswhere (see picture).
Now comes the part where you need to "scale" sounds of different loudnesses this way. Generally, as you pointed out, that would require a different correction for each loudness. But: if, in addition, the equal-loudness curves are equi-distant at all frequencies, then the absolute loudness level is no longer relevant and the difference between 2 curves which are say 10 phon apart at 1000 hz is the same no matter their absolute level. This is key for implementing a "static" equalization and, according to the ISO model, - but not to the original FM model ! - a reasonable assumption.
Unfortunately, the live vs. playback loudness ratio is content and recording dependent and cannot be trivially computed out of the recording. Secondly, the recording contais some equalization which is difficult to factor in. If you solve this problems, you can, in theory, implement accurate loudness correction.
That is precisely because IMHO nobody did it right so far, for the reasons I outline in the previous posts. You need to know the the original-to-target ratio and assume that the curves are equi-distant (they are in the ISO model, which is pure-tone based, but who knows how accurate that is for music ?)
Attachments
Last edited:
I'm not particularly interested in the F-M curves being applied dynamically based on the signal itself. I realize the impracticality of such a system.
I am now focused on building a loudness circuit which is tied to the volume control knob and which can be easily tuned to a particular system. Once set up, a certain amount of "loudness" is applied based on the position of the volume control.
As someone else said, this would not take into account the recording level, source level, etc. but it would certainly be "good enough" for my aging ears and tastes.
I've decided to build this myself. It'll likely be built as a series of prototypes which progress from a simple, passive, variable control to something integrated into my pre-amp and tied to the physical volume control.
Best of luck to those who choose to try something similar in the digital realm.
I am now focused on building a loudness circuit which is tied to the volume control knob and which can be easily tuned to a particular system. Once set up, a certain amount of "loudness" is applied based on the position of the volume control.
As someone else said, this would not take into account the recording level, source level, etc. but it would certainly be "good enough" for my aging ears and tastes.
I've decided to build this myself. It'll likely be built as a series of prototypes which progress from a simple, passive, variable control to something integrated into my pre-amp and tied to the physical volume control.
Best of luck to those who choose to try something similar in the digital realm.
I think bzfcocon did a great thing in that explanation of the problem, possible ways forward and pragmatic simplifications that could be done to get a "good enough" result. Thanks.
But I may be missing something but doesn't Dolby and/or Audassy implement this in some reasonable fashion? (Dynamic Volume or similar) Given they have a reference level with movie soundtracks to work with, recorded music is more messy in that regard..
Are these DSP solutions bad (having no experience of them myself) or perhaps good (enough) but undocumented and proprietary..?
/Niclas
But
I may be missing something but doesn't Dolby and/or Audassy implement this in some reasonable fashion? (Dynamic Volume or similar) Given they have a reference level with movie soundtracks to work with, recorded music is more messy in that regard.. Are these DSP solutions bad (having no experience of them myself) or perhaps good (enough) but undocumented and proprietary..?
/Niclas
I've been looking at this issue only recently, I also thought "there must be something out there already" but I do not know of any particularly clever DSP based solution (better than convetional loudness circuits), which does not mean that it does not exists.
Would be great if you could provide some details/links.
Would be great if you could provide some details/links.
Allright, I've been looking at the Dolby Volume, there a technical paper, but it's not really so technical as one woul wish for. Clearly, it does exactly the sort of dynamic corrections we've been talking about above.
Dolby Volume for Playback Volume Control
http://www.dolby.com/us/en/technologies/dolby-volume-tech-paper.pdf
No idea how good or bad it is, especially for music, they do claim that they try not to affect dynamics. It should be available in some higher-end AV equipment. I'd still wish for a static eq solution though.
Some similar stuff should be available from Audissey. According to some internet discussions. Dolby should be better.
Dolby Volume for Playback Volume Control
http://www.dolby.com/us/en/technologies/dolby-volume-tech-paper.pdf
No idea how good or bad it is, especially for music, they do claim that they try not to affect dynamics. It should be available in some higher-end AV equipment. I'd still wish for a static eq solution though.
Some similar stuff should be available from Audissey. According to some internet discussions. Dolby should be better.
Nice that you explained loudness compensation, as practiced by pre-amp manufacturers. But my feelings are hurt that you thought it was taking me 50 years to understand the basic math manipulation.
But now I see you are beginning to notice there are problems with your model.
You are assuming there is a little man (or woman or except where prohibited by law, an LGBT person) in your your head who says, "Gosh, who turned down the bass on that soft recording?"
Shouldn't the little person say, "Ummm, that sounds just perfectly right for Beethoven's symphony played at a great distance away."
The starting point of my puzzlehood is a comparison to the vision system where "size-distance invariance" usually (but not always) operates. When you see your car 200 feet away across a parking lot, you never say, "OMG, somebody shrunk my car to just one inch long!!!!"
Correspondingly for hearing, when you hear Beethoven at 200 feet (or played very softly while others sleep in your house), shouldn't you say, "That sounds just great albeit from 200 feet away."?
And that's just the start of my puzzlehood.
Ben
Last edited:
I didn't mean to hurt your feelings. To be honest, the explanation was more for myself. Despite basic math, it's not so easy to intuitively grasp the model for practical purposes.
The problems I have outlined are not with the model, but with the lack of information input into the model (live event loudness).
OK, to me it seems you not puzzled about the "how", but about the "why". Now you're no longer talking about equalizing loud vs. soft differently - I'd say we're one step forward.
You analogy is not entirely accurate because there is more to (auditive) distance perception than just relative loudness, there are also spatial cues. Also, the visual and auditive systems are not working similarly. But let's accept it for the moment.
The point is: 200 feet away is not a good listening position, nor is it an enjoyable place to look at your car. To enjoy the look of your car, you can 1. go closer to your car or 2. use a lens device (telescope) that corrects your view so your car appears to be closer.
Translated in audio, option 1 means turn the volume up so it sounds as if you're in the concert seat and option 2 means use a loudness correction that acts like a lens/telescope to make things appear closer.
Now, option 2 is not "the real thing" for either visual or auditive. You still can't walk around your car just as you won't exactly get the "in the concert seat" feeling. But it's still a lot better than 200 feet away !
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.