Harman Curve Compliance

[mk05]

I'm not technically inclined as most of this forum, and I mainly lurk to retain some information and glean some wisdom from time to time. So forgive me if this has been covered already, if this is a stupid question, and/or any negative perceived tone (I am just curious).

One thing that I can't seem to understand here on ASR is this "Harman Curve" for headphones, and the requirement for "compliance" to it as some benchmark. I totally understand a need to have a benchmark to rate things, but I am curious on the why. A quick search tells me that a Harman Curve is

"a target frequency response curve in audio engineering, specifically for headphones and loudspeakers, that aims to simulate how most listeners perceive sound in typical listening situations."

From my understanding of ASR, everyone here has been preaching about a flat response being the benchmark for loudspeakers, as they are supposed to reproduce sound as neutrally as possible. So why does the standard differ for headphones (although Harman seems to want to apply this to loudspeakers as well)? Light search into the Harman Curve tells me that it is based on supposed aggregate preference. So if the Harman curve is allegedly preference, who says this is what I will like? Why do I need to get a flat response loudspeaker, but a curvy looking response headphone? Everyone here trashes certain brands of loudspeakers as not good when they exhibit a "house EQ" or a Harman Curve compliance, but when headphones have them, they are perfect. This does not make any sense to me.

Arguably, isn't Harman Curve just pushing the industry into making skewed sounding speakers? Thank you in advance, really interested!

 

[digitalfrost]

The harman curve for speakers is what you would measure in-room with "good" speakers. Good speakers in the sense that they measure flat in an anechoic (no echos, i.e. no room) environment and have smooth directivity, ideally. If you put these into a room, they will have more bass and less high frequency as the air will absorb some high frequency energy and higher frequencies are more easily reflected due to their shorter wavelength.

Now when you listen to good speakers in a good room, the sound that you hear with your ears, is neither flat nor the harman speaker target curve, because humans ears are extremely nonlinear.

Equal-loudness contour - Wikipedia

en.wikipedia.org

You can see our ears are extremely sensitive around the 3khz region and less so in the bass, especially if the bass is quiet. Now if you put on headphones, the headphones must simulate what you would hear, if you had good speakers in a good room. But because the room is not there, we must fake the room.

An extreme variant of this would be IEMs, since here you remove even the outer part of the ear. See also https://en.wikipedia.org/wiki/Head-related_transfer_function

 

[Bruce Morgen]

you remove even the outer part of the ear.

That sounds painful. Is anesthesia available?

 

[markanini]

So why does the standard differ for headphones

It's prudent to educate yourself on basic differences between different systems before questioning standards. With speaker measurements, the goal is to characterize them as free-field omnidirectional radiators to capture their true performance. They project sound into a space, interacting with the environment and the listener, so including variable room acoustics makes consistent, comparable data impossible. For headphones, in-situ conditions are simulated instead, mimicking human head and ear anatomy to attempt to predict their performance for an average listener. Their operation depends on physical coupling to the ears, so this can’t be skipped.

 

[unpluggged]

It's prudent to educate yourself on basic differences between different systems before questioning standards. With speaker measurements, the goal is to characterize them as free-field omnidirectional radiators to capture their true performance. They project sound into a space, interacting with the environment and the listener, so including variable room acoustics makes consistent, comparable data impossible. For headphones, in-situ conditions are simulated instead, mimicking human head and ear anatomy to attempt to predict their performance for an average listener. Their operation depends on physical coupling to the ears, so this can’t be skipped.

That still does not mean the Harman OE and IE phone statistical preference curves should be regarded as the benchmark target FRs for all and any case. On the opposite, many people (me included) do not like them so much and do not consider them the absolute benchmarks and even remotely as neutral-sounding. At best, they are just another target FRs that can be used to design phones with specific sound profile that will appeal to people who like this kind of signature.

It does not mean that diffuse-field targets or some alternative IEM targets proposed recently sound any worse. It does not also mean that phones must or even should strive to emulate the sound of a speaker in a room in the first place, as Harman targets do.

 

[DVDdoug]

Speakers should be flat on-axis in an anechoic chamber. Off-axis they won't be flat. That includes studio monitors used to mix & master music.

When you put those real-world speakers/monitors in a room (or studio) where the off-axis sounds bounce around are mixed with the direct sound, you tend to get a downward-sloping curve. The mixing & mastering engineers make it "sound right" under those conditions and that's what we're trying to reproduce at home.

Headphones sound different from speakers and flat headphones sound different from flat speakers. Also with speakers you can feel the bass in your body (if it's strong enough) and of course, you lose that with headphones. The Harman curve approximates the sound (the frequency response) that you perceive in a room (or studio) with good speakers. (Although it was actually developed as a preference curve.)

You may have a different preference. If you have a pair of headphones and you know how they measure (how much they deviate from the Harman curve), that can help you decide if you like a little more, or a little less, bass than the standard, etc.

 

[markanini]

That still does not mean the Harman OE and IE phone statistical preference curves should be regarded as the benchmark target FRs for all and any case.

Did I make that conclusion?

 

[Madlop26]

It's prudent to educate yourself on basic differences between different systems before questioning standards. With speaker measurements, the goal is to characterize them as free-field omnidirectional radiators to capture their true performance. They project sound into a space, interacting with the environment and the listener, so including variable room acoustics makes consistent, comparable data impossible. For headphones, in-situ conditions are simulated instead, mimicking human head and ear anatomy to attempt to predict their performance for an average listener. Their operation depends on physical coupling to the ears, so this can’t be skipped.

Hmm can we try to simplify so is easier to understand?
Research has showed that a speaker flat response in is what most people will like, and headphones being a different system needs adjustment to equal that , hence the Harman curve..

 

[solderdude]

From my understanding of ASR, everyone here has been preaching about a flat response being the benchmark for loudspeakers, as they are supposed to reproduce sound as neutrally as possible.

Yes, speakers that measure 'flat' in anechoic circumstances AND have good directivity, a wide enough frequency response, low in distortion at higher SPL sound good in most of the 'average listening room'.
The Harman curve for the speaker at the listening position in a room differs from the flat response measured anechoically. You get a certain boost (usually peaky/lumpy) in the lows (hence the rise in the lows) and the higher frequencies roll-off gently.

So 'flat measuring speakers' will have a bit more bass and less higher frequencies at the listening position. This was explored by Harman thus Harman curves. There are other curves too.
It measures flat when using a 'normal' mic. Speakers usually are not tested with a HATS.

So why does the standard differ for headphones (although Harman seems to want to apply this to loudspeakers as well)?

1: The speakers in a headphone are not in a large room at the same angle as speakers would be.
2: The speakers in a headphone are in a small (somewhat) sealed small chamber, very close to or against the ears.

The angle thus differs.
The distance differs
There is no 'changing of the sound caused by the room'.
There is a direct coupling to the ear.
The signals from the left ear do not reach the right ear.
Measurements are done with a fake ear that modifies the incoming sound.


All of these aspects (especially the last one) means that what is picked up by the mic inside the 'fake eared/ear-canal' head picks up a totally different (20dB) signal.
The goal thus is to 'compensate' for the alterations the HATS makes.
On top of that we want the headphones to have a similar tonal balance of 'flat speakers in an ideal listening room'.
That is the 'target'.

With HATS measurements we thus have to 'undo' the changes the ear+canal makes + the target (more lows, gently rolled off treble) a speaker does.
So totally different 'corrections' to obtain the same tonal balance.
With speakers we also 'feel' (tactile) the lowest bass frequencies which we don't with headphones so some correction for the lack of tactile feel also alters the target.

Light search into the Harman Curve tells me that it is based on supposed aggregate preference.

Yes, the evaluations are based on many experiments with people and thus preference plays a role.
The good part is that most people prefer 'good tonal balance' and generally only differ in the amount of bass (for various reasons)

So if the Harman curve is allegedly preference, who says this is what I will like?

No one does. you nor I nor others have to... most, however, do prefer that tonality.
Harman curve is what is preferred by the majority of people.
That means there still is a substantial group that prefers more or less bass and more or less treble.
There are other standards too, some are somewhat applicable as well.

Arguably, isn't Harman Curve just pushing the industry into making skewed sounding speakers? Thank you in advance, really interested!

The Harman research was started in order to educate Harman engineers. It just grew and as it is the most substantial body of research on this topic it is kind-of bombarded as 'a standard'. There are more standards.
Amir had to choose a standard and the Harman target is the logical choice here.
All measured headphones are to that standard so ... easy to compare.

 

[unpluggged]

Did I make that conclusion?

No, just to clarify.

 

[mk05]

...
1: The speakers in a headphone are not in a large room at the same angle as speakers would be.
2: The speakers in a headphone are in a small (somewhat) sealed small chamber, very close to or against the ears.

The angle thus differs.
The distance differs
There is no 'changing of the sound caused by the room'.
There is a direct coupling to the ear.
The signals from the left ear do not reach the right ear.
Measurements are done with a fake ear that modifies the incoming sound.


All of these aspects (especially the last one) means that what is picked up by the mic inside the 'fake eared/ear-canal' head picks up a totally different (20dB) signal.
The goal thus is to 'compensate' for the alterations the HATS makes.
On top of that we want the headphones to have a similar tonal balance of 'flat speakers in an ideal listening room'.
That is the 'target'.

With HATS measurements we thus have to 'undo' the changes the ear+canal makes + the target (more lows, gently rolled off treble) a speaker does.
So totally different 'corrections' to obtain the same tonal balance.
With speakers we also 'feel' (tactile) the lowest bass frequencies which we don't with headphones so some correction for the lack of tactile feel also alters the target.

Yes, the evaluations are based on many experiments with people and thus preference plays a role.

Thank you so much, it clarifies some things, but makes me wonder about some other things. I definitely understand the idea behind these endeavors (Harman curve creation) is to try to help people foremost, and formulate some sort of standardization. My questions, as one of the members alludes, are simply too mundane probably for this forum, but I'd still like to venture if allowed.

Because I have never seen how headphones were measured, I *assumed* that there would be a microphone aimed at the headphone speaker, much like a loudspeaker projects to a mic. Apparently that is not the case and it is rather more like this:

Firstly, I would contend that the Harman curve is still a preference EQ since Harman itself says is a curve preferred by aggregate. Nothing wrong with that, but then it is not objectively neutral? Who gets to claim that it "feels" neutral? Or is the data saying that the Harman curve signature would be what would present as a flat response if it were measured in front of a room with a mic pointed at it? (probably butchered how things are measured, but please work with me)

Secondly, let's say that I understand the premise of headphones being a different application since ears are on the side of the head and go through an ear canal, etc as the picture above indicates, which requires those specific mics and whatnot. If this is the case, why are we pointing one mic at a loudspeaker in front of it, then? Why not put that whole entire head assembly on the Klippel, and measure a loudspeaker in that fashion like headphones, since we have 2 ears and aren't turning our head to direct one ear at one speaker, but rather perpendicularly to both? Wouldn't that yield data that is more relevant to a real human head interaction with sound, much like the contention for why Harman curve exists for headphones? I'm not sure if I am explaining myself very well.

I know these questions must sound absolutely ridiculous, but I've wondered this for so long and I might as well just expose myself as a casual.

 

[FTB]

Everything you need to know (I guess) : https://danishsoundcluster.dk/wp-content/uploads/2022/11/Olive_DSD_2022.pdf

 

[NTK]

A summary of the origin of the Harman headphone target curve:

• We normally listen to music in a room. Research by Dr Floyd Toole and Sean Olive, et al., have shown what good loudspeaker so should be like.
• Therefore, they put a HATS (head and torso simulator), similar to but not the same as the setup below (the picture is from a different study), in the prime listening position in their reference listening room to measure the response of a good loudspeaker setup.
  
  
  
  
  
  
  
• They then used the same HATS to measure headphones and apply EQ to the headphones to match the response to the response of the loudspeakers, and have people listen to the headphones EQ to the Harman curve, no EQ, and and EQ to other curves. After some tweaking to the curve based on listening tests, the outcome of their 2018 study is that 64% of listeners preferred the latest Harman target curve at the time.
  
  
  
  
  

 

[markanini]

Hmm can we try to simplify so is easier to understand?
Research has showed that a speaker flat response in is what most people will like, and headphones being a different system needs adjustment to equal that , hence the Harman curve..

Speakers propagate soundwaves through a room. Headphones are attached directly to your head and use physical tricks to sound vaguely like speakers in a room. Is this simple enough?

 

[solderdude]

Firstly, I would contend that the Harman curve is still a preference EQ since Harman itself says is a curve preferred by aggregate. Nothing wrong with that, but then it is not objectively neutral?

No it is not objectively neutral as that is not possible. It is a compromise based on many factors. Found 'curves' are averaged. Also music is used and not all music is recorded 'neutral'.
Furthermore tonal balance is SPL dependent.

So NO any curve is not objectively neutral. It is a compromise, an average based on input data.
Talking about the headphone curve (that was revised a couple of times)

Who gets to claim that it "feels" neutral?

Every test subject. Average them and you end up with ... well ... an average. When the majority of the test subjects hover around the found average then you have a target that suits the majority of listeners. And that is exactly what the Harman target is.

Secondly, let's say that I understand the premise of headphones being a different application since ears are on the side of the head and go through an ear canal, etc as the picture above indicates, which requires those specific mics and whatnot. If this is the case, why are we pointing one mic at a loudspeaker in front of it, then?

With speakers we want to know the response and the dispersion pattern.
Once you have a reference speaker and set it up in a specified 'room' and you measure the response with a HATS at the listening position then you know what response you need for a headphone (when using the same fixture).
But then.... you are missing some aspects like crossfeed, missing tactile feel so even when the measurements would be the same you still would not get the same 'sound' as the body and crossfeed also play a role in determining the sound.

Why not put that whole entire head assembly on the Klippel, and measure a loudspeaker in that fashion like headphones, since we have 2 ears and aren't turning our head to direct one ear at one speaker, but rather perpendicularly to both? Wouldn't that yield data that is more relevant to a real human head interaction with sound, much like the contention for why Harman curve exists for headphones? I'm not sure if I am explaining myself very well.

When you want to accurately measure sound you need a microphone that is known to measure flat. Why would anyone want to use a microphone with dips and peaks that vary with distance, elevation, angle (that can be well over 20dB off) which are nearly impossible to correct for every angle, distance and elevation ?
That's why speakers are measured with microphones and not with HATS.

Headphones are on the head. There is a pinna that influences the sound (angle, elevation, distance) which differs from headphone to headphone, there is seal, positioning etc.
This makes Klippel measurements reliable and accurate and headphone measurements a crap shoot that is somewhat repeatable on the same fixture.

Speakers you measure with a microphone.
Acoustics of a room you measure with a HATS or microphone or combination.
Headphones you measure on a (suited for the task as not all HATS are suited) test fixture or HATS.

Headphones simply are not speakers close to ones head.
They have a thing in common... they are both used for listening to music but in a very different way.

 

[Overseas]

Useful presentation of H Curve history and alternative curves:

https://www.howtogeek.com/what-is-the-harman-curve-and-why-do-people-like-it-so-much/

 

[digitalfrost]

OP. This video should also explain better why the curves look like they do:

 

[GREG K]

Hmm can we try to simplify so is easier to understand?
Research has showed that a speaker flat response in is what most people will like, and headphones being a different system needs adjustment to equal that , hence the Harman curve..

Yes, the H curve for headphones is intended to "sound like" flat speakers in a room.

 

[solderdude]

Yes, the H curve for headphones is intended to "sound like" flat speakers in a room.

For about 60% of the people, the other 40% prefer a different tonal balance.
Also it is based on a 'pretty good and big listening room with good speakers'.
The issue is that most people do not have those speakers nor that room and any 'reference' they have with their speakers in their room may well be very different from that of a headphone on a specific test fixture with a highly averaged target.