At which sound pressure levels are loudspeakers calibrated to achieve a flat perceived loudness curve?

[Endibol]

(good) speakers are designed to be anechoically flat when measured by a microphone. Not perceptually flat. The curve of equal loudness perception is never flat at any SPL.

The real variable that matters is "at what spl is the content mixed/mastered?" since almost all content is going to be altered by EQ, compression, etc before it is released. And the answer there is usually about 80-85dB, it depends on the audio engineers, the type of content/standard they're mixing for(tv/film have real standards, music doesn't), etc.

Given that amount of bass/treble is also a matter of preference, as long as you listen around 80dB you're probably hearing more or less what was intended. If you listen significantly quieter, say 70dB, you likely need to add some bass boost via EQ or a room correction target curve.

I mean subjectively flat.

 

[fpitas]

I mean subjectively flat.

Right, that's a problem for the mixing and mastering guys, and you, once you place them in your room.

 

[Endibol]

I understand this, but which curve would you employ to not corrupt the fidelity (assuming we don't mind wrecking linear phase to achieve the effect)?

You can never know what the correct shape is, because:
a) Each curve is a slightly different shape (and they are an average of many people, so each individual is different!)
b) The shape would be relative to the "correct" listening level for the recording. Bob Katz and his K-System assumes that monitoring level is at 85dB, so we could use that. But we know that each recording is potentially slightly different.
c) The curves (and the equivalent ISO versions), don't exist - they are like isobars. In other words, you can't switch from curve 7 to curve 8, because there are an infinite number of curves in between, each as slightly different shape and getting more extreme the quieter the listening level is.
d) There is often "gain-riding" by the mixing engineer and mastering engineer and sometimes they apply compression, which means very few recordings have louder quiet sections than would happen in real life, thereby reducing the Fletcher-Munson curve!

Ultimately, there is no way to reliable create a Loudness button which doesn't mess up phase and which actually corrects for lower listening levels. That doesn't mean people don't benefit from a bit of bass and treble boost at lower listening levels, but they are not accurately correcting for equal loudness and are messing up phase, so Loudness buttons always distort.

I understand that dynamic loudness adjustment is both difficult to achieve and undesirable. My point is that by definition the loudspeaker manufacturer implements a certain spectral balance by the design of the fixed cross over networks and the woofer/tweeter caracteristics. This means that out of the infinite amount of the Fletcher-Munson curves (phon-curves) one is selected (albeit unintentionally) that corresponds to a certain phon-level. Which phon-curve do they pick? Is there an industry standard?

 

[ahofer]

there's no reason you can't have a system that continuously changes the FR depending on output level.

The RME DAC does this, with user-customized boost and roll-out.

 

[GXAlan]

loudspeaker manufacturer don't do this anymore it has been decades. a flat response is targeted.

This isn't what you see with measurements of B&W 800-series speakers though?

 

[Sancus]

I mean subjectively flat.

I don't really get what you mean. Like I said, speakers are designed to be flat when measured by a microphone in an anechoic chamber. Equal loudness contours don't come into it at all.

Any compensation for equal loudness contours is embedded in the music itself due to the mixing engineer's EQ choices, or in some(rare) cases built into certain electronics or active speaker systems.

To elaborate, I think there a bunch of misunderstandings in this thread:
1) Microphones are just reverse speakers. They have sensitivity and directivity, same as a speaker. They DO NOT record what a human would hear if they were sitting in the same spot as the microphone. That is why microphone choice and placement are so important, and why nobody is recording music with a single microphone. They're almost always using at least 3 if not many more.

2) The frequency balance/spectrum of a recording is altered arbitrarily by MANY factors between the recording and playback steps. At the recording step, all the attributes of the microphones and the recording engineer's setup of them alters the sound. Then again at the mixing and mastering steps. Even for some weird types of classical music where they skip the mixing step, they are still mixing the audio from several microphones and the placement, angle, and microphone choice also affects things.

3) To even know the SPL that's being recorded/produced at each step, calibration must be done. Microphones record arbitrary loudness deepending on their sensitivity and interaction of their directivity with the source sound. The only way you'd know that the microphone will record "80dB SPL" is if you calibrated it for a specific source, angle, distance, etc. And again at the mixing/mastering steps, the audio engineer has to have their speakers calibrated for a specific SPL.

Because of point 3, especially for music, there's no real way to know what SPL level the music was balanced for. You can make an educated guess, and that's it. Precision is not possible. Read Dr. Toole's book, especially the part that talks about the circle of confusion.

In order for accurate playback all the way through from recording to consumer, every single step would need to be standardized, from microphones used in recordings, to speakers used for mixing/mastering, to speakers used by the end user for playback. And you would need information on what the calibrated SPL was as well. Parts of this standardization do exist for Film/TV content, but not a full chain, and pretty much none of it exists for music.

At the end of the day adjusting for equal loudness contours is a subjective process for the end user. A compensation system can HELP, but even if you built a perfect compensation system(you would need a calibrated active system that has measured the exact SPL at the listening position), you would still need to tell that system what SPL the content was intended to be played at. And that's not information you usually have.

 

[kemmler3D]

I mean subjectively flat.

"subjectively flat" is a thing for headphones, but not speakers. The reason you need a compensation curve in headphones to reach "subjectively flat" is because the headphones bypass the torso, head, and parts of the pinnae, making the frequency response at the eardrum not "flat" compared to playing the same audio on a speaker.

Flat speakers are both objectively and subjectively flat on their own.

Loudness compensation becomes a thing if you are really concerned that your playback level is too far from "reference", such that you lose subjective (fletcher-munson) flatness relative to what was heard at the performance or in the studio. That is an aspect of the signal, not the loudspeaker, so flat speakers are still flat even if you're listening "too quietly".

 

[-Matt-]

This isn't what you see with measurements of B&W 800-series speakers though?

Yes, the "bat ears" "v" profile much maligned in these parts could very well function as a sort of baked in loudness correction.

I.e. It could make it more satisfying when listening to music at lower levels, because the boosted bass and trebel get you closer to the intended sound balance when the music was mastered loud.

 

[ahofer]

I.e. It could make it more satisfying when listening to music at lower levels, because the boosted bass and trebel get you closer to the intended sound balance when the music was mastered loud.

Yeah, the live reference is always louder. Those instruments/ensembles are typically playing at levels that would be intense in your living room. So Hifi reproduction is an exercise in scaling down.

 

[Cote Dazur]

So the Fletcher-Munson curve is NEVER encoded in the signal, so amplifiers with a loudness button distorts the signal and a speaker that tries to correct for it will ALWAYS be wrong

Thank you for your post, makes perfect sense to me on a subject that I was always confuse with.

 

[Endibol]

I don't really get what you mean. Like I said, speakers are designed to be flat when measured by a microphone in an anechoic chamber. Equal loudness contours don't come into it at all.

Any compensation for equal loudness contours is embedded in the music itself due to the mixing engineer's EQ choices, or in some(rare) cases built into certain electronics or active speaker systems.

To elaborate, I think there a bunch of misunderstandings in this thread:
1) Microphones are just reverse speakers. They have sensitivity and directivity, same as a speaker. They DO NOT record what a human would hear if they were sitting in the same spot as the microphone. That is why microphone choice and placement are so important, and why nobody is recording music with a single microphone. They're almost always using at least 3 if not many more.

2) The frequency balance/spectrum of a recording is altered arbitrarily by MANY factors between the recording and playback steps. At the recording step, all the attributes of the microphones and the recording engineer's setup of them alters the sound. Then again at the mixing and mastering steps. Even for some weird types of classical music where they skip the mixing step, they are still mixing the audio from several microphones and the placement, angle, and microphone choice also affects things.

3) To even know the SPL that's being recorded/produced at each step, calibration must be done. Microphones record arbitrary loudness deepending on their sensitivity and interaction of their directivity with the source sound. The only way you'd know that the microphone will record "80dB SPL" is if you calibrated it for a specific source, angle, distance, etc. And again at the mixing/mastering steps, the audio engineer has to have their speakers calibrated for a specific SPL.

Because of point 3, especially for music, there's no real way to know what SPL level the music was balanced for. You can make an educated guess, and that's it. Precision is not possible. Read Dr. Toole's book, especially the part that talks about the circle of confusion.

In order for accurate playback all the way through from recording to consumer, every single step would need to be standardized, from microphones used in recordings, to speakers used for mixing/mastering, to speakers used by the end user for playback. And you would need information on what the calibrated SPL was as well. Parts of this standardization do exist for Film/TV content, but not a full chain, and pretty much none of it exists for music.

At the end of the day adjusting for equal loudness contours is a subjective process for the end user. A compensation system can HELP, but even if you built a perfect compensation system(you would need a calibrated active system that has measured the exact SPL at the listening position), you would still need to tell that system what SPL the content was intended to be played at. And that's not information you usually have.

@Sancus Thank you (and others in this thread) for their extensive response to my question. Let's see whether I now understand it correctly with a thought experiment:

Suppose we offer a constant SPL-sweep over the audible bandwidth to a perfectly calibrated microphone. In that case the output of this microphone is a constant voltage across the frequency-sweep. This signal is fed via a perfectly linear amplifier to a loudspeaker. This (ideal) loudspeaker has been designed in such a way that an input voltage sweep will create a constant SPL (say at 1m distance in an anechoic room). If we now use the same microphone to record the reproduced sound sweep and show the microphone output signal on an oscilloscope we measure a constant voltage. In this simple case there is no medium (like CD, flac-file etc.) part of the signal chain.

Theoretically we could have stored the output voltage from the microphone on a medium like CD using perfectly linear devices without interference by a human being.
In practice when we deal with music, humans are needed to process the signal before it is stored onto CD.

Therefore a human being, the sound engineer, is introduced. This is a subjective step, since we are now depending on the ears/perception of this particular individual. The sound engineer will tweak the voltage levels to be stored on the CD for the low and high frequencies relative to the mid-frequencies based on his/her individual "built-in" Fletcher-Munson curve. So this is done based on a subjective opinion of what sounds "natural" or "desired" at the loudness level that has been chosen to work with. For this and other loudness levels the CD-listeners at home may need to adjust the "equalizer" on their amplifier to make it sound pleasant for their own ears and brain.

Do I get it now? Thanks!

 

[TSB]

You are at a concert. A cello is playing very quietly. Because of the Fletcher-Munson curve, you perceive the the middle tones as louder than the higher and lower notes. It plays much louder, suddenly the lower and higher tones are clearer.

Next to you is a microphone. It doesn't suffer from the curve! It records faithfully to a machine, which also doesn't suffer from the Fletcher-Munson curve. You listen to the result on a perfectly flat set of speakers. The quiet bits, just like the concert has quieter lower and upper tones, because your hearing has the Fletcher-Munson curve. Whilst the louder bits, just like the concert have clearer upper and lower tones.

So the Fletcher-Munson curve is NEVER encoded in the signal, so amplifiers with a loudness button distorts the signal and a speaker that tries to correct for it will ALWAYS be wrong.

I believe this is partially correct, however:

- Playback levels are (usually) lower than the original concert
- The Fletcher-Munson curves are a family of curves, a different one for each playback level
- Hence the Fletcher-Munson curve you use to interpret the lower level reproduced sound is different from the one you use to interpret the reference level sound at the concert
- Thus: reproduction sounds significantly different from the live concert unless played at the reference level of the concert
- Add to this the work of the recording engineer, who tries to make the recording sound good at a different reference level (not the concert level) , using specific equipment available in the recording studio (which presumably, does not have loudnes compensation)

So we have quite a 'circle of confusion'..
- Maybe some kind of loudness compensation can help in some cases
- Clearly not possible to define a single solution that works for all recordings

I find most classical recordings hard to listen to at lower levels, much to the annoyance of some

 

[MaxwellsEq]

There's no circle of confusion. The mixing engineer cannot encode a loudness curve into mix. Ends.

 

[TSB]

There's no circle of confusion. The mixing engineer cannot encode a loudness curve into mix. Ends.

The players in the concert implicitly encode a loudness curve during the concert by their artistic decisions during the concert.

Their decisions depends on what they hear, which is dependent on loudness curves.

If you ask them to play at lower volumes, they will make (perhaps unconsciously) different decisions to ensure the balance of the sound.

Similarly, a recording engineer will make decisions based on what he hears at reference level. What he hears depends on loudness curves.

 

[RayDunzl]

I ask myself: at which sound pressure level do the loudspeaker manufacturers adjust the balance between the woofer and tweeter?

If listening "critically", I aim for around 80dB SPL average, 100dB peaks.

That's a comfortably loud level, and corresponds to the flatter area of the Fletcher Munson graph.

Speakers don't change their tune with different volume.

I figure the players (or the mixer) adjusts levels in the recording for playback audibility.



MartinLogan reQuest at various SPL, measured at the listening position, with a "flat" EQ, below.

Ambient noise messes with the low level bass response readings.

1/12th octave smoothing

Psychoacoustic averaging:

"Flat" EQ creates (for me) a measurable similarity between the speaker response and the raw signal source.

If playing at lower levels, I'm not annoyed by any lack of bass. Still "seems" flat to me, as any low frequency sound rolls off at lower levels.


I tried adding bass according to the FM curve, didn't care for the result, anyway.

 

[MaxwellsEq]

If you ask them to play at lower volumes, they will make (perhaps unconsciously) different decisions to ensure the balance of the sound

Really? Do you mean you really believe musicians play ppp but boost the lower notes and higher notes, but not the middle notes? Seriously?

 

[DonH56]

Dolby reference level for audio calibration is 75 dB IIRC.

 

[MaxwellsEq]

One calibration approach is Bob Katz' K-System https://www.aes.org/technical/documentDownloads.cfm?docID=65f

 

[TSB]

Really? Do you mean you really believe musicians play ppp but boost the lower notes and higher notes, but not the middle notes? Seriously?

Yes.

Conductors spend significant effort to belance the sound of their orchestra at different levels. For this they use their ears, hence a loudness-specific Fletcher curve.

 

[MaxwellsEq]

Yes.

Conductors spend significant effort to belance the sound of their orchestra at different levels. For this they use their ears, hence a loudness-specific Fletcher curve.

They do not.

The value between ppp and fff is dictated in the score. Beethoven did not tell pianists to hit the middle keys gently and the lower and upper keys hard during ppp. When you are taught an instrument, nobody discusses Fletcher-Munson curves. Conductors never give instructions to musicians to play to a certain Fletcher-Munson curve. If you are sat in an orchestra, but your instrument group is at rest, and another instrument is playing, when they play ppp, you have to strain to hear the music and your own Fletcher-Munson curve is active for that sound level; when they play fff it's much clearer.

That's the whole point of dynamics in a live orchestra!