• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Headphones and the Harman target curve

Yorkshire Mouth

Major Contributor
Joined
Apr 22, 2020
Messages
1,254
Likes
1,153
Location
God's County - Yorkshire
But in practice, what does it really matter? Does anyone stick hard and fast to the EQ settings, for eg, Oratory1990 to Harman, or in fact plug those filters in first then tweak. I almost always lower the bass filters and often some of the others. How "accurate" the Harman target is only ever really becomes an issue if the only filters you can use are pre made convolved ones which cant be adjusted by the user.

But its interesting to refresh how we got to Harman. A more interesting one is how we get from the original HP curve to the one now.

I'm probably not going to put this very well, but here's my feeling.

I think this view...further legitimises debate, and gives us a fresh view.

May I offer an example. AKG K371s. Between 20hz & 100hz it's heavier on bass than Harman by c.4dB. If Harman is 6dB too bass heavy to start with, that makes for a 10dB overshoot.

In comparison, the Sennheiser HD600/650s slope from 100hz down to 20hz, starting at Harman, and ending up 11dB below at the very end.

Using that measure, the HD600s are closer to 'great speakers in a great room' (GSIAGR) than the AKGs.

From now on, when I look at Amir's headphone reviews, here's a thing I'll have in mind. When I look at what is usually his second graph (frequency response deviation from target), anything below 100hz which is within 6db of the blue line, I think needs to be considered 'within range'. And at the top, anything from around 2.5khz up to 20khz which averages 3dB above is also 'within range'.

And again, regarding your comments on applying and experiment with EQ, I think it's good to know we can use it, but I suspect most of us would like to have to use it as little as possible, due to added issues such as distortion. Using the GSIAGR target curve, you have to add a hell of a lot less EQ at the bottom end, which is clearly preferable.

I think you're right, JimBob. I think we need an exact Harman curve less and less, and more of a range. For certain, it'd be wrong to say the Harman curve is empirically 'wrong'. But surly it's equally wrong to say that one curve is 'right', or should be our target.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
I'm probably not going to put this very well, but here's my feeling.

I think this view...further legitimises debate, and gives us a fresh view.

May I offer an example. AKG K371s. Between 20hz & 100hz it's heavier on bass than Harman by c.4dB. If Harman is 6dB too bass heavy to start with, that makes for a 10dB overshoot.

In comparison, the Sennheiser HD600/650s slope from 100hz down to 20hz, starting at Harman, and ending up 11dB below at the very end.

Using that measure, the HD600s are closer to 'great speakers in a great room' (GSIAGR) than the AKGs.

From now on, when I look at Amir's headphone reviews, here's a thing I'll have in mind. When I look at what is usually his second graph (frequency response deviation from target), anything below 100hz which is within 6db of the blue line, I think needs to be considered 'within range'. And at the top, anything from around 2.5khz up to 20khz which averages 3dB above is also 'within range'.

And again, regarding your comments on applying and experiment with EQ, I think it's good to know we can use it, but I suspect most of us would like to have to use it as little as possible, due to added issues such as distortion. Using the GSIAGR target curve, you have to add a hell of a lot less EQ at the bottom end, which is clearly preferable.

I think you're right, JimBob. I think we need an exact Harman curve less and less, and more of a range. For certain, it'd be wrong to say the Harman curve is empirically 'wrong'. But surly it's equally wrong to say that one curve is 'right', or should be our target.

Perhaps I can save you some time:
1) Different listeners prefer different amounts of bass. It varies based on age, gender, and country of residence. This is demonstrated in Harman research. There is also variation in treble preference.
2) Rather than represent #1 as a shaded range vs freq, Harman still chooses to represent a "target curve" as a thin line and suggests that headphones should follow the curve and listeners NEED to adjust bass/treble to their liking. Fine. Semantics. Interestingly Harman represents the target room response for loudspeakers as a shaded range of amplitude vs freq. So much for consistency.
3) Harman's research on headphone curves isn't as definitive as you might think. They took a pair of AKG headphones and eq'd them to the measured FR of various other HP models and subject that to preference testing. Some of the supposedly top rated HP models using this method sound like absolute garbage. And even then, the correlation between overall headphone FR and listener prefs was good but not great.
4) What all of this really means is that HP measurements are an incomplete substitute for actual listening at this point in time if you want to really know how a pair of HP's sound to you.
 
Last edited:

Yorkshire Mouth

Major Contributor
Joined
Apr 22, 2020
Messages
1,254
Likes
1,153
Location
God's County - Yorkshire
I don’t disagree with any of that.

I think possibly it’d just help if we drop ‘the target’ or ‘our target’.

It’s clearly one option among many, and I don’t believe it has any more validity than many of the alternatives, other than an averaged personal preference which demonstrably does not mirror the GSIAGR starting point.

We wouldn’t accept that sort of deviation because of personal preference in any other area we discuss here.
 

BrEpBrEpBrEpBrEp

Active Member
Joined
May 3, 2021
Messages
201
Likes
245
I believe the green, dotted line on Slide 10 is the ‘pre-preference’ curve.

Comments there reflect those linked to earlier, that Harman is basically:

1 - From 250hz down to 20hz Harman rises by around 6dB, and that should be flat, or at most a more modest +2dB.

2 - Harman's 'peak' at 3-4khz is around 2-2.5dB too low, and remains around 2-2.5dB low for the remainder.

Not necessarily - that dotted line is the loudspeakers EQed to flat. However, good speakers in a good room exhibit a downwards sloping response - boosted bass and recessed trebled. They just repeated the speaker preference tests with headphones, which, as they said, showed the need for boosted bass and recessed treble. See paragraph one of conclusions.

1622487533780.png
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
I don’t disagree with any of that.

I think possibly it’d just help if we drop ‘the target’ or ‘our target’.

It’s clearly one option among many, and I don’t believe it has any more validity than many of the alternatives, other than an averaged personal preference which demonstrably does not mirror the GSIAGR starting point.

You're certainly entitled to your opinion that the Harman headphone curve is is no more valid than other curves out there. However, the science indicates that it does correlate with listener preferences, and published papers help us understand how much correlation there is. Are you able to share any other headphone FR curves that have been demonstrated by research to correlate with listener preferences (or a suitable gold standard, if you don't like "listener preferences")?

We wouldn’t accept that sort of deviation because of personal preference in any other area we discuss here.

I'm not sure what this is referring to, perhaps you can clarify. Again, the Harman room curve gives a range of values vs. frequency, and it's pretty well accepted.
 

ReaderZ

Addicted to Fun and Learning
Joined
Apr 14, 2020
Messages
618
Likes
414
You're certainly entitled to your opinion that the Harman headphone curve is is no more valid than other curves out there. However, the science indicates that it does correlate with listener preferences, and published papers help us understand how much correlation there is. Are you able to share any other headphone FR curves that have been demonstrated by research to correlate with listener preferences (or a suitable gold standard, if you don't like "listener preferences")?



I'm not sure what this is referring to, perhaps you can clarify. Again, the Harman room curve gives a range of values vs. frequency, and it's pretty well accepted.

You can't claim "the science indicates that it does correlate with listener preferences" unless the paper is published by someone with no tie or funding from Harman and the paper has been peer reviewed. So far what I have seen is results published by Harman and it's not enough.
 

Yorkshire Mouth

Major Contributor
Joined
Apr 22, 2020
Messages
1,254
Likes
1,153
Location
God's County - Yorkshire
You're certainly entitled to your opinion that the Harman headphone curve is is no more valid than other curves out there. However, the science indicates that it does correlate with listener preferences, and published papers help us understand how much correlation there is. Are you able to share any other headphone FR curves that have been demonstrated by research to correlate with listener preferences (or a suitable gold standard, if you don't like "listener preferences")?



I'm not sure what this is referring to, perhaps you can clarify. Again, the Harman room curve gives a range of values vs. frequency, and it's pretty well accepted.

I’m not sure how to respond to that.

The Harman curve is, unequivocally based on measurement of preference.

The pre-preference curve is based on pure data.

In any other area of science, in any other area we discuss here, there would only be one choice to make.

We wouldn’t measure distortion, and then ask how much distortion people prefer. We’d just say that distortion does but faithfully and accurately reproduce the original. That would be the end of the matter.

I find it strange that people feel a room introduces issues which support a move towards adding bass and reducing treble. Most (not all) of the issues with rooms come from walls artificially amplifying bass and muffling treble.

If people listen to a GSIAGR and don’t like it, then the problem lies with them. In much the same way that people who prefer Beats headphones or massive subwoofers in the back of the car cannot have their opinions attributed to accurate reproduction.

Someone once said that they appreciate the next man’s opinion on religion and politics, but only in as much as they appreciate his opinion that his wife is gorgeous and his kids are smart.

Harman test listeners’ opinion on how much bass and treble they prefer over GSIAGR fall into the same category for me.

GSIAGR, that’ll do. If you want something different, that’s your opinion. I’m just want the original recreated.

Not the original + more bass, ‘cos dats cool innit.
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
Do you think listening to speakers in an anechoic chamber is a suitable reference for your audio consumption? If so the green doted curve is a good match. Be aware that's not how the vast majority of the audience consumes media, and not a space audio engineers work in.
Lipinski-Test-e1509774048673.jpeg
 
Last edited:

Yorkshire Mouth

Major Contributor
Joined
Apr 22, 2020
Messages
1,254
Likes
1,153
Location
God's County - Yorkshire
Do you think listening to speakers in an anechoic chamber is a suitable reference for your audio consumption? If so the green doted curve is a good match. Be aware that's not how the vast majority of the audience consumes media, and not a space audio engineers work in.

That’s not what the green dotted line is.

The green dotted line is GSIAGR. NOT a good speaker in an anechoic chamber.

If it were not, I’d agree with you.
 

BrEpBrEpBrEpBrEp

Active Member
Joined
May 3, 2021
Messages
201
Likes
245
That’s not what the green dotted line is.

The green dotted line is GSIAGR. NOT a good speaker in an anechoic chamber.

If it were not, I’d agree with you.
The green dotted line on slide 10 is GSIAGR, EQed to be dead flat - not a normal listening setup. This is comparable to the response of GSI an anechoic chamber. Compared to the green dotted line, GSIAGR means more bass and less treble - i.e. comparable to the Harman target.
 

Yorkshire Mouth

Major Contributor
Joined
Apr 22, 2020
Messages
1,254
Likes
1,153
Location
God's County - Yorkshire
There’s a ‘science’ to measuring preference. This is not the same as, or equal to, the science of measuring accurate and faithful transparency.

Had Harman wanted the latter, they could have had it. VERY easily.

Either:

1 - Take the curve from GSIAGR and simply point to that.

or

2 - Take test curve and played it back over headphones to trained listeners in the same room, asking them to switch between headphone listening and speaker
listening, to see if the curve for the headphones accurately reproduces the sound in the room.

The moment you deviate from this and ask for a preference, you’ve lost it. You’re going for headphone sales, not accurate reproduction. Don’t get me wrong, that has its own validity, but it’s not in line with absolutely everything else that’s done here at ASR.

In short, scientifically measuring preference is different to scientifically measuring accuracy.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
You can't claim "the science indicates that it does correlate with listener preferences" unless the paper is published by someone with no tie or funding from Harman and the paper has been peer reviewed. So far what I have seen is results published by Harman and it's not enough.

Not sure why you thought it was okay to make up rules here. For starters, the Harman research in question has been published in a peer-reviewed journal (JAES). Can you think of a higher-tired journal that would be suitable for this type of work? I can't. Secondly, plenty of research is privately funded.

Unless, of course, you're insinuating that the Harman-funded work published by Drs. Toole and Olive in JAES were falsified in some way. Perhaps you can clarify whether that's what you're saying, because it's a different conversation.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
There’s a ‘science’ to measuring preference. This is not the same as, or equal to, the science of measuring accurate and faithful transparency.

How would you define "accuracy" or "faithful transparency" in a measurable way? You mention "science," and science requires observability or measurability.


Either:

1 - Take the curve from GSIAGR and simply point to that.

or

2 - Take test curve and played it back over headphones to trained listeners in the same room, asking them to switch between headphone listening and speaker listening, to see if the curve for the headphones accurately reproduces the sound in the room.

It's been done. And the theoretical FR curve measured at the ears that represents the FR of GSIAGR doesn't translate into the headphones that are accurate or preferred. Part of the reason is that a simple FR curve (and btw, you haven't even defined where you're taking the measurement from, i.e. in the canal, at the external auditory meatus, etc.) does not fully describe for what is heard. In reality, sound waves are striking the pinna at all sorts of different angles of attack, each with various millisecond delays and differences in amplitude with respect to one another. How, pray tell, could you possibly represent all of that with a single amplitude vs. frequency curve?

The moment you deviate from this and ask for a preference, you’ve lost it. You’re going for headphone sales, not accurate reproduction. Don’t get me wrong, that has its own validity, but it’s not in line with absolutely everything else that’s done here at ASR.

Again, you need to define how one should measure "accuracy." If you can't observe or measure it, it ain't science.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
I’m not sure how to respond to that.

The Harman curve is, unequivocally based on measurement of preference.

The pre-preference curve is based on pure data.

And unfortunately, the so-called "data" doesn't completely describe the reproduced sound that the ear is exposed to. That's the problem. And that's why the pre-preference curve isn't necessarily the most accurate. See my post above. If you don't fully understand the data you've collected, you're not necessarily going to draw the right conclusion.

We wouldn’t measure distortion, and then ask how much distortion people prefer. We’d just say that distortion does but faithfully and accurately reproduce the original. That would be the end of the matter.

I can tell that you don't have a lot of experience designing natural science experiments. So, YES, you COULD characterize how audible and objectionable distortion is by employing a group of listeners capable of providing a preference score, and then varying the amount of distortion present in a sound clip. The hypothesis would be that greater % THD (using THD as an example) would result in lower preference scores.

I find it strange that people feel a room introduces issues which support a move towards adding bass and reducing treble. Most (not all) of the issues with rooms come from walls artificially amplifying bass and muffling treble.

That sounds like an explanation straight out of the Crutchfield catalog.

If people listen to a GSIAGR and don’t like it, then the problem lies with them. In much the same way that people who prefer Beats headphones or massive subwoofers in the back of the car cannot have their opinions attributed to accurate reproduction.

Again, that's way oversimplified. The experience of bass and its perceived volume differs substantially between headphones and loudspeakers in a room. Many over-ear phones are flat between 20Hz and 100Hz, yet their bass is subjectively "weaker" than loudspeakers playing flat from 20Hz to 200Hz in a room. This is a completely different phenomenon than purposely boosting bass.
 
Last edited:

ReaderZ

Addicted to Fun and Learning
Joined
Apr 14, 2020
Messages
618
Likes
414
Not sure why you thought it was okay to make up rules here. For starters, the Harman research in question has been published in a peer-reviewed journal (JAES). Can you think of a higher-tired journal that would be suitable for this type of work? I can't. Secondly, plenty of research is privately funded.

Unless, of course, you're insinuating that the Harman-funded work published by Drs. Toole and Olive in JAES were falsified in some way. Perhaps you can clarify whether that's what you're saying, because it's a different conversation.

Please link to JAES paper on the latest harman headphone curve.
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
The green dotted line on slide 10 is GSIAGR, EQed to be dead flat - not a normal listening setup. This is comparable to the response of GSI an anechoic chamber. Compared to the green dotted line, GSIAGR means more bass and less treble - i.e. comparable to the Harman target.
You don't even have to take if from Harman. Take a look at Amirs speaker measurements. All well designed speakers show a sloped in-room response, not straight. That's what is being accounted for regarding the the difference between the black curve and and the green dotted curve.
 
Top Bottom