• 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

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
Help what? And who? Are you referring to a way to eliminate the Circle of Confusion?
That might be a side effect, yes. Sound engineers could note down how their gear deviates from some known reference and as such (having same information about their gear) user could easily make adjustments towards better reproduction. Also it could serve to reduce all the "house sound of brand" bollocks.

What you describe has already been done. AKG N90Q. It did auto EQ on your ears and had built in bass and treble tone controls. Overall, it sounded GOOD but NOT EXCELLENT. It also got discontinued without a follow up model, suggesting that it was a failed experiment from a profitability standpoint.
Yet that just shows there wasn't a market for that between AKG customers, mostly. Still doesn't mean that boutique brands wouldn't get away with selling the idea. Obviously it's more profitable to keep on selling different visions of sound in different models instead of standarizing.

Well, for starters, when talking about human subject preferences, there is always going to be variability in your measured data - with very few exceptions. That's like saying a standard male is 5'-10" and 70kg and designing everything around those parameters.

Also, a simple, smoothed, 2-d amplitude vs frequency curve (such as a "target curve") is not a complete characterization of what is "heard" by a set of headphones. In reality, what is being heard is a polar frequency response that is dependent on angle of incidence to the ear axis with micro time delays caused by reflections inside the ear cup, which represents a 4th dimension time). For this reason, you can't "EQ" an HD600 into an HD800S (and I'm pretty sure countless people have tried and were sorely disappointed).
Yes and no, since in the end even HD800 won't reproduce the stereo delay close to what speakers do. In the end, it's all perceived as one sound unless the delay is over X ms (read up precedence effect), X being between 50 and 100.
 

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
That might be a side effect, yes. Sound engineers could note down how their gear deviates from some known reference and as such (having same information about their gear) user could easily make adjustments towards better reproduction. Also it could serve to reduce all the "house sound of brand" bollocks.

Sounds fantastic. Standardization is a good thing when it comes to manufacturing and process improvement. But the thing is, there usually a strong motivation (usually financial) to standardize and improve in those industries. I just don't see that same compelling need from manufacturers in the audio industry. Perhaps from an academic standpoint but thats it.

Yes and no, since in the end even HD800 won't reproduce the stereo delay close to what speakers do. In the end, it's all perceived as one sound unless the delay is over X ms (read up precedence effect), X being between 50 and 100.

That is simply not true. A 1khz sine wave at x dB amplitude striking the pinna at 30 degs is not necessarily going to be perceived the same as if it were aligned at 0 degs, even if time aligned. Furthermore, a given measured frequency response in the canal for human A wearing a set of headphones is going to be different from the measured frequency response at the canal for human B wearing those same headphones. The anatomical differences in human pinna shape are responsible. Finally, within the chamber formed by a headphone cup around the ear, there are addiditonal standing waves and frequency response variations introduced that will also introduce variation. And I'm just getting started. Headphones acoustics bring a whole different set of considerations. Yes an average preferred curve can be determine empirically, but it's very simplified represention of the sound being presented to the ear. And if you don't believe me, buy a pair of the headphones whose curve, when reproduced on the AKG test phones in the Harman study, had the highest predicted pref score. They sound absolutely horrendous. Amir reviewed them and he agrees (although he was more diplomatic than me in his writeup.)

For over the ear headphones, the measured frequency response doesnt tell the whole story.
 

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
Sounds fantastic. Standardization is a good thing when it comes to manufacturing and process improvement. But the thing is, there usually a strong motivation (usually financial) to standardize and improve in those industries. I just don't see that same compelling need from manufacturers in the audio industry. Perhaps from an academic standpoint but thats it.
Being versed in IT and not much more, there is usually no real effort towards standarization if "standarizing" doesn't mean "pushing our own product as a standard", which is ridiculous but that's how it works. Breaking the mould is problematic. Sennheiser keeps on producing HD600 for years and years and that's the way to introduce an actual reference point. Of course I agree that audio market is not interested in lessening the snake-oil potential.

That is simply not true. A 1khz sine wave at x dB amplitude striking the pinna at 30 degs is not necessarily going to be perceived the same as if it were aligned at 0 degs, even if time aligned. Furthermore, a given measured frequency response in the canal for human A wearing a set of headphones is going to be different from the measured frequency response at the canal for human B wearing those same headphones. The anatomical differences in human pinna shape are responsible. Finally, within the chamber formed by a headphone cup around the ear, there are addiditonal standing waves and frequency response variations introduced that will also introduce variation. And I'm just getting started. Headphones acoustics bring a whole different set of considerations. Yes an average preferred curve can be determine empirically, but it's very simplified represention of the sound being presented to the ear. And if you don't believe me, buy a pair of the headphones whose curve, when reproduced on the AKG test phones in the Harman study, had the highest predicted pref score. They sound absolutely horrendous. Amir reviewed them and he agrees (although he was more diplomatic than me in his writeup.)

For over the ear headphones, the measured frequency response doesnt tell the whole story.
I think there are two separate things You're mixing up together: measuring the FR at eardrum or inside ear canal at v.small time intervals would tell us the whole story, but how to produce a specific FR for the given anatomy is a whole another story which might require varying the construction. Setting a common curve brings us closer to solving the first (given far-field consideration, which AFAIR is good when modelling generalized sound emission w/o modelling the angle of incidence) problem - what FR to target (for a given modelled set of angles), but might not necessarily help how to solve the latter. Still, also the whole headphone-based sound spatial virtualization (where aforementioned issues are more important) is available now and seems to be pretty well-done.
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
I don't really understand why we're having the circular discussions here. For me it boils down to two seperate things, there's Circle of Confusion which is purely related to lack of standards in the recording studios in terms of how "flat" the speakers were in the recording studio where the music was made, and then the second variable is the Target Curve of choice in headphones.

Circle of Confusion
The best starting point is Anechoically Flat Speakers in your room, then........well you just use a Tilt Tone Control to broadly alter tonality on a per track / per music albumn basis. Like for me Red Hot Chilli Peppers record everything overly bright, so you'd use Tilt Tone Control to darken the speakers to compensate. The same broad Tone Control Filter applies to headphones too in order to combat this Circle of Confusion variable. (to be honest I often can't be bothered to get up to activate the Tone Control Filter on a per song or per albumn basis, lol!)

Headphone Target Curve
This is a seperate issue to the previous point, this is always innacurate to a greater or lesser extent depending how much your anatomy is different from the average anatomy (dummy head) that the Target Curve has been based on, and it's the same issue no matter which dummy head or which type of target curve is used (although some better than others). The only way to be as accurate as possible is to use Smyth Realiser or Impulcifier Project to create your own personalised Target Curve that takes your anatomy into account. Creating your own Target Curve though is only as accurate as the speakers & room you're basing it on when referring to the Impulcifier Project, so there's that variable additionally. Regarding the Harman Headphone Target Curve, the same rules apply being it based around an average anatomy, but it's a proven researched curve that makes a lot of sense, and it's attempting to replicate good speakers in a good room (done through actual measurements) and then some preference study added onto that. Harman Headphone Target is the best we got unless you create your own personalised Target Curve.

Therefore in terms of accuracy Headphones are imperfect, whereas speakers are not (as long you have perfect Anechoically Flat Speakers in a perfectly treated room, and organised in an equilateral triangle with the speakers & listener at each of the corners of that triangle). I'm not sure why we're having these circular conversations in this thread the past few days.....it just boils down to these two bolded points above.
 
Last edited:

DeLub

Active Member
Joined
Jun 16, 2020
Messages
135
Likes
178
Location
The Netherlands
I'm not sure why we're having these circular conversations in this thread the past few days.....it just boils down to these two bolded points above.
We're having these discussion because these two points don't justify a bass bump of 6dB in the Harman curve for any anatomy...
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
We're having these discussion because these two points don't justify a bass bump of 6dB in the Harman curve for any anatomy...
That doesn't seem to be the only sticking point, but on that point we know that's part of the preference study part of the process.....and the theory of why that might be quite as high is due to the fact that you don't have tactile bass from headphones that you would otherwise have in speakers, therefore the amount of bass in headphone listening has to be somewhat higher to make up for the lack of tactile bass, that's the most accepted theory that I've found for that, and to me it makes sense. I see nothing wrong with that conclusion, because by definition headphone listening is a different experience to speakers, so that's just one of the ways in which the subjective experience of bass can be equalled between headphones & speakers.
 

Jimbob54

Grand Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
11,063
Likes
14,694
Seems to me we are spoilt for choice regarding eq for headphones. The bass bump that most hp need to hit Harman is too large to my ears but that doesn't really matter much, does it?

Unless one is going down a permanent hardware mod route don't we just adjust the one or 2 bass filters according to our taste or preferred music styles?

As for accuracy, after a certain point, largely chasing rainbows, surely?
 
Last edited:

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Seems to me we are spoilt for choice regarding eq for headphones. The bass bump that most hp need to hit Harman is too large to my ears but that doesn't really matter much, does it?

Unless one is going down a permanent hardware mod route don't we just adjust the one or 2 bass filters according to our taste or preferred music styles?

As for accuracy, after a certain point, largely chasing rainbows, surely?
I agree re the bass point you bring up. Re Accuracy, then for headphones I agree, but for speakers I don't.......but for headphones with a personalised Target Curve through Impulcifier Project (if done well) and Smyth Realiser then I think headphones can be quite accurate as it's tailored to your own anatomy, but headphones always innacurate to some extent if you are using any target curve based on a dummy head.
 

Jimbob54

Grand Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
11,063
Likes
14,694
I agree re the bass point you bring up. Re Accuracy, then for headphones I agree, but for speakers I don't.......but for headphones with a personalised Target Curve through Impulcifier Project (if done well) and Smyth Realiser then I think headphones can be quite accurate as it's tailored to your own anatomy, but headphones always innacurate to some extent if you are using any target curve based on a dummy head.

For speakers it does seem a done deal, as close as you can get to that gentle downward slope in room. And again, some tilt and /or steady shelves after that to mood /style.

And let's not of course forget, if your first headphone is something a fair distance from Harman (let's say a dt990, EQ to Harman is going to sound WEIRD. I think that first try puts a few off.
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
For speakers it does seem a done deal, as close as you can get to that gentle downward slope in room. And again, some tilt and /or steady shelves after that to mood /style.

And let's not of course forget, if your first headphone is something a fair distance from Harman (let's say a dt990, EQ to Harman is going to sound WEIRD. I think that first try puts a few off.
Yeah, definitely, so easy to get used to the sound signature of some very strange headphones if you don't have reference speakers to compare against or experience with other headphones....yep and then potentially good sound can sound very strange at first once your mind is burnt into a headphone with a strange frequency response. I think it's probably fair to say that a headphone with an off-the-wall frequency response is not gonna luck it out to match someone's HRTF (anatomy) to produce an accurate sound, so it would have been just a case of mind getting burnt in to a crazy weird frequency response. Not that Harman is accurate sound, as it's based on a dummy head, so can't possibly be perfectly accurate, but on average for the population it's likely to be closer to accurate sound than a random selection of headphone frequency responses out there.
 
Last edited:

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,554
Likes
1,701
Location
California
Being versed in IT and not much more, there is usually no real effort towards standarization if "standarizing" doesn't mean "pushing our own product as a standard", which is ridiculous but that's how it works. Breaking the mould is problematic. Sennheiser keeps on producing HD600 for years and years and that's the way to introduce an actual reference point. Of course I agree that audio market is not interested in lessening the snake-oil potential.

Exactly. What practical incentive is there for companies to get together and agree on a single target curve at the risk of alienating their current loyal following and eliminating brand differentiation?

I think there are two separate things You're mixing up together: measuring the FR at eardrum or inside ear canal at v.small time intervals would tell us the whole story, but how to produce a specific FR for the given anatomy is a whole another story which might require varying the construction. Setting a common curve brings us closer to solving the first (given far-field consideration, which AFAIR is good when modelling generalized sound emission w/o modelling the angle of incidence) problem - what FR to target (for a given modelled set of angles), but might not necessarily help how to solve the latter. Still, also the whole headphone-based sound spatial virtualization (where aforementioned issues are more important) is available now and seems to be pretty well-done.

I hear you, and at the same time, when taking FR measurements of various headphones at the canal using a standardized pinna on validated measurement equipment, FR does not explain all the variability in blinded listener preferences when analyzed statistically. In fact, the correlation is even less strong than that for loudspeakers, which we already know is less than perfect.

In other words two manufacturers can produce headphones that have a more or less identical FR at the canal and these two headphones would still likely sound different and differ how they are preferred.

The idea that there is a single 2D FR curve that, when achieved will always result in the most preferred sound by all listeners is a fantasy. And because of that I personally see no reason to standardize to one.
 

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
The idea that there is a single 2D FR curve that, when achieved will always result in the most preferred sound by all listeners is a fantasy. And because of that I personally see no reason to standardize to one.
The idea is to have a standard curve that is more or less the least far away from any single preference to have to adjust the least for each.
 

Swoquix

New Member
Joined
Oct 29, 2021
Messages
2
Likes
0
I've read several times here that the Harman Curve predicts the FR that most people will like in headphones. Just wondering if there is an operational definition of the word most in the research? I'm thinking 85-90% or better when I think of most but I'm wondering what the research states.

edited to add "the FR"
 

Jimbob54

Grand Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
11,063
Likes
14,694
I've read several times here that the Harman Curve predicts the FR that most people will like in headphones. Just wondering if there is an operational definition of the word most in the research? I'm thinking 85-90% or better when I think of most but I'm wondering what the research states.

edited to add "the FR"
The research details the %. And states the bandings that deviate from it. Search for it on Google or the multitude of threads about it on here. From memory its around 70%
 

Sean Olive

Senior Member
Audio Luminary
Technical Expert
Joined
Jul 31, 2019
Messages
329
Likes
3,004
The research details the %. And states the bandings that deviate from it. Search for it on Google or the multitude of threads about it on here. From memory its around 70%

In this paper https://www.aes.org/e-lib/browse.cfm?elib=20289 we looked at segmentation of listeners based on headphone preference and found 64% of the test population fell into a cluster of listeners who preferred the target. The other two segments like the Harman target but preferred models that had slight more or less bass.
1635540226683.png
 

Jimbob54

Grand Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
11,063
Likes
14,694

Swoquix

New Member
Joined
Oct 29, 2021
Messages
2
Likes
0

GaryH

Major Contributor
Joined
May 12, 2021
Messages
1,348
Likes
1,804
In this paper https://www.aes.org/e-lib/browse.cfm?elib=20289 we looked at segmentation of listeners based on headphone preference and found 64% of the test population fell into a cluster of listeners who preferred the target. The other two segments like the Harman target but preferred models that had slight more or less bass.
View attachment 162052

Hi Sean, I have a question about this study. As I understand, these classes were determined through cluster analysis of preference ratings given to a set of headphones, so the classes are partly dependent on the predefined frequency responses of this set, which I would think could skew the results. As most headphones (at least at the time of the study) either had boosted (mid/upper) bass, or lacked bass (most open-backs), as seems apparent from the below graphs, is it not unsurprising 3 classes of preference were found (the other being the large majority of Harman target lovers of course)?

index.php


I believe the bass control method of adjustment in previous studies is more representative of true preference as it does not rely on a predefined selection of headphone bass shelves, instead the listener is free to choose any bass level they like. Is it possible (or have you already done) a cluster analysis of bass preference from these method of adjustment studies to see if this concurs with the results from the headphone preference study?
 

mga2009

Active Member
Joined
Nov 21, 2019
Messages
165
Likes
84
Hi,

Can anyone share the Harman curve for headphones (or any other recommended TC) in TXT format so I can use it with REW?

Thanks in advance.
 
Top Bottom