7 Comments

It’s hard for me to believe it’s been seven years since researchers from Harman International presented the landmark paper “The Relationship between Perception and Measurement of Headphone Sound Quality” at the 2012 Audio Engineering Society Convention in San Francisco -- the first paper in which the company presented what became known as the “Harman curve,” the target frequency response that average listeners would like best. When I first read that paper, I assumed it would quickly revolutionize the headphone business. As a headphone reviewer, I knew that the various headphones and earphones then on the market often exhibited wildly different sounds -- even among different models in a company’s line -- which indicated that knowledge was lacking or simply being ignored.

Sean Olive

Yet we didn’t see a surge in headphones claiming to use the Harman curve -- not even from Harman companies AKG, JBL, and Harman Kardon, although other manufacturers had quietly confided to me that they were basing their designs on the Harman curve. In fact, the first set of passive headphones I’ve received from a Harman company using the Harman curve came across my desk only last month: the AKG K371s -- and the only way I knew that they were voiced within 1dB of the Harman curve was by reading a Facebook post from Harman senior fellow Sean Olive.

It looks, though, like the dam is finally bursting. In a presentation last month to the Los Angeles chapter of the AES, Olive highlighted several other headphones and earphones designed along the lines of the Harman curve (although not, as best I can tell, marketed as such). The earphones include Samsung Galaxy Buds; JBL Live 200s, Live 500s, Live 650s, and Reflect Flows; and AKG N5005s. Headphones, for now, seem to include only the AKG N700 NCs, K371s, and K361s, but we can expect more. As Olive told me later in an e-mail, “Basically all new AKG headphones are designed to Harman target, and for the past year JBL has followed it but with 2dB extra bass below 125Hz.”

Olive’s presentation detailed Harman’s ambitious research into headphones, which has since 2012 resulted in 19 papers, one patent (and three more pending), and a one-click routine for the SoundCheck audio measurement suite that uses the Harman curve to predict listener preference. The effort started with the realization that, as Olive put it, “There were standards for diffuse-field and free-field headphone response, but no one was following them so there must have been something wrong.

201911 akg twist

“At the time, our marketing department was telling us that we should duplicate the response of Beats headphones, because those were the best sellers,” he continued. But Harman’s researchers had already evaluated those headphones in blind tests, and they found them to be unpopular among their listeners. “So we told them they should duplicate Beats’ marketing instead,” he said.

The researchers’ idea was that if headphone designers knew what measured response best suited the largest number of listeners, the designers could then tailor their products’ frequency responses to that target. This would be more efficient than “shooting in the dark” by building headphones, putting them through listening tests (which, if you want them to be blind, are much more complicated with headphones than they are with speakers), then repeating the process until most of the listeners are pleased.

Harman’s first effort involved a blind test of six over-ear headphones, followed by measurements of those response curves to see which response pleased the most listeners. Subsequent projects solicited the judgment of hundreds of listeners around the world and measurements of hundreds of different headphones.

Sean Olive

The results of all that effort were Harman curves for earphones and over-ear headphones. But as Olive suggested in his above comment about JBL headphones, it’s not a “one size fits all” target. His presentation identified three potential groups to which manufacturers can target their headphones.

“Harman curve Lovers”: This group, which constitutes 64% of listeners, includes mostly a broad spectrum of people, although they’re generally under age 50. They prefer headphones tuned close to the Harman curve.

“More Bass Is Better”: This next group, which makes up 15% of listeners, prefers headphones with 3 to 6dB more bass than Harman curve below 300Hz, and 1dB more output above 1kHz. This group is predominantly male and younger -- the listeners JBL is targeting with its headphones.

“Less Bass Is Better”: This group, 21% of listeners, prefers 2 to 3dB less bass than the Harman curve and 1dB more output above 1kHz. This group is disproportionately female and older than 50.

According to Olive, his group still has some more research into headphones to do, but they’re starting to wrap it up and anticipate moving on to new projects. We don’t yet have enough information to know if the Harman curve will result in greater sales, millions more happy listeners, and better standardization of headphone evaluation. I’ve learned from listening to several headphones and earphones that come close to the Harman curve that it is, at the very least, an excellent baseline for performance. Headphones and earphones may, for various reasons, deviate to some extent from the Harman curve. But if their measured response is far from the Harman targets, listeners and reviewers should at least question why, and the manufacturer should be able to respond with a plausible rationale based on research and testing.

I expect we’ll encounter manufacturers and reviewers who simply claim, “I listened to some Harman curve headphones and I don’t think they sound very good.” Audio enthusiasts will then have to decide whether they trust the conclusion of a single person, determined through casual, sighted testing, or the conclusion of years’ worth of research conducted with hundreds of listeners in carefully controlled blind tests. I know which way I’ll go.

. . . Brent Butterworth
This email address is being protected from spambots. You need JavaScript enabled to view it.

Say something here...
Log in with ( Sign Up ? )
or post as a guest
People in conversation:
Loading comment... The comment will be refreshed after 00:00.
  • This commment is unpublished.
    Mauro · 1 years ago
    @Brent Butterworth Eheh, I didn't want to put the bug in your ear!

    It's something I have been wondering about as well.
    I guess that being the target audience pretty young they have competitive prices and frequency response is good also because they are usually active designs..so worth considering for the budget-minded fascia.

    About the specific headphone, I would say that from 60 Hz to 7kHz is pretty spot-on. Below 60Hz drops quickly, but I guess it might be expected given the dimension of the driver, right?
    At around 8Khz there is a little hint of "sss" sound, that make voices a bit lipsy, but not a big deal honestly.

    My quest into open headphone is moving to Philips Fidelio X2 I found at around 100€, very entertaining! They have a huge soundstage! A pity some bloat at very low frequences and a quite predominant "sss" sound at top, that makes treble a bit artificial..

    Waiting to get the LCD-1 I have just ordered on your suggestion!
  • This commment is unpublished.
    Brent Butterworth · 1 years ago
    @Brent Butterworth But I really should, huh? I'll look into it ...
  • This commment is unpublished.
    Brent Butterworth · 1 years ago
    @Dustin No, sadly. I would like to find out, but Sean is my only source (unless I get to measure a sample), and I'm not sure when the guidelines he cited were put into place.
  • This commment is unpublished.
    Brent Butterworth · 1 years ago
    @Bernd Unfortunately your anonymity prevents me from confirming your research work; I hope you'll be willing to share your identity, and I think it's appropriate, necessary and courteous to do so when you're criticizing another researcher's work.

    I've not stated or implied that the Harman Curve is the "holy grail," and neither has anyone from Harman, to my knowledge. As I've stated in reviews, I haven't necessarily found that Harman Curve headphones sound best, but they all sound good, to me and the other listeners I work with. That's why I said it's a useful baseline.

    And in my opinion, that's what we want: a headphone that sounds good. While it's common to suppose that listeners have radically different tastes, and would thus need all sorts of different headphones, I don't know of any research that confirms this supposition, and I do know of research that refutes it. In fact, Harman did exactly this research, and found that with 238 test subjects in three countries, preferences were generally consistent, “regardless of the listeners’ experience, age, gender and culture.” Saying that "practicable studies are always limited" is not a substantive refutation of this or any other research.

    While I'm guessing this is not your intent, what you're saying reminds me of what I hear from many high-end audio writers and audiophiles. To wit: Science is so complicated and uncertain that we can't really know anything for sure, so in the absence of absolute certainty, I'll just believe whatever feels right to me. Or as my friend Dennis Burger once summed up this view, "Science is sometimes wrong, so homeopathy works."

    And then what do we have to judge audio products by? Totally subjective reviews from people who mostly fit the description in my previous post. I'll pass. We have good science here and I intend to use it and promote it.

    It doesn't sound to me like you're intentionally advocating the "faith-based" approach to audio that most high-end publications take, but as I see it, the end result is the same -- unless you can present something better than the existing research. And I certainly encourage you to do so, and I promise to write about it if you do!
  • This commment is unpublished.
    Bernd · 1 years ago
    @Brent Butterworth Thanks for you reply, Brent. Actually I've conducted and partially published a fair share of listening tests myself, at university and later in product development, so I'm speaking from experience as well.

    But I believe we both agree on the difficulties when trying to achieve reliable, repeatable, and valid listening tests and comparisons. They are impaired by short acoustic memory, choice of stimuli, listener fatigue, conflict between isolation of factors vs. interdependence of factors (esp. with headphones: wearing comfort, sealing, usage of EQs or presets, heavy pre-processing/sweetening by streaming services, listening situation and intention, ...); sometimes a fully double-blind design just isn't feasible; etc.

    And since most headphones are targeted at the worldwide market, then this worldwide population should be correctly represented in the study; the same goes for the various listening scenarios, target groups (defined by more than age, gender and nationality) etc. for which the products are designed and marketed.

    Next is the difficulty of measuring what you actually hear; standardized pinnae and couplers only enable repeatability, but don't necessarily replicate what arrives at your ear drum. Sometimes, narrow details in the response are just as important as the general FR envelope, but can't be represented in "cookbook recipe" curves. In all this, we aren't even talking about driver type and size, transient behaviour, linearity etc. yet, which also influence preference.

    When considering all this, as well as all the factors I mentioned in my first post, it should become clear that we are dealing with a really really complex problem, if we aim at fully "solving" the question of how a headphone should sound. Achieving validity (representativeness and meaningful methods) as well as reliability (significance and repeatability) despite this complexity, is what leads me to the admittedly vague estimation of "thousands" of listeners. I didn't say that every listening test has to be very large, but if you strive for universal answers regarding headphones, that's what would be needed. And yes, often the standard deviation or confidence of listening test results doesn't improve when increasing the number of repetitions or listeners, which then indicates that the test design is flawed or that the subject matter is very difficult to determine.

    So don't get me wrong, I'm all for the scientific method, we don't need to debate here about the details of how to apply it to preference testing. I'm just saying that practicable studies are always very limited, given the vast overall complexity of the subject, and therefore the range of valid conclusions is very limited as well. Treating Sean's results as universally valid would be disregarding those limits of validity.

    My ultimate point is that we shouldn't demand all new headphones to converge towards the Harman curve - there are many good reasons (some of them mentioned before) for implementing a different voicing. Seeking *the* one (or very few) "holy grail" curves seems somewhat similar to the gut-felt "this is the ultimate"-claims of the high-end gurus which you understandably say you have no respect for.
  • This commment is unpublished.
    Brent Butterworth · 1 years ago
    @Bernd The listener base for "A Statistical Model That Predicts Listeners' Preference Ratings of In-Ear Headphones" was 71 listeners, ages 20 to 59, and 82% male, roughly half trained and half untrained listeners. In "Factors That Influence Listeners' Preferred Bass and Treble Levels in Headphones," they used 249 listeners, with groups in Michigan, California, Germany and China, ages 18 to 70, 89% male. Sean admits that he'd like to have more female listeners, but as someone who's assembled listening panels for 25+ years, I know how challenging it is to recruit female listeners for audio tests.

    Can you present evidence that a scientifically solid audio study requires thousands of listeners? That claim would invalidate most listening-based research. Can you present evidence that a wide variety of listening curves is needed? That contradicts the results of the Harman studies, and TMK there are no competing studies that have disproved their contentions.

    It's common in audio for people to dismiss measurements, listening tests, etc. as invalid because they find some flaw or inconsistency, real or perceived. But the alternative, especially in high-end audio, is not studies with more listeners or a different methodology. It's one guy, often with spotty knowledge of audio research, a tenuous (or non-existent) connection to music performance and production, a set of ears with many decades of wear and tear on them, and sometimes without even a solid grasp of engineering, declaring that he knows what music is supposed to sound like and he's created a product that "sounds like music" or some other vague claim that can't be proven or disproven. Then a reviewer, who likely also fits the above description, and who uses no evaluation process deserving of the term "method," provides flowery verbiage to support the claims.

    There are high-end designers who really do know the research and engineering, and who create terrific products (Paul Barton, Andrew Jones and Laurence Dickie all come right to mind, but of course there are many, many others), but the gut-feel "process" I describe above is probably at least as common. This is exactly the kind of thing that scientific research is intended to combat, and for which I personally have no respect.

    My attitude is, if you want to criticize scientific research, cite references to back up your claims, and invest the time and resources to find a better answer.
  • This commment is unpublished.
    Bernd · 1 years ago
    @Bernd (the comment function replaces the number three with hearts, for some reason, testing again: 3
  • This commment is unpublished.
    Bernd · 1 years ago
    Brent, you should mention the listener base used in Olive's studies. For someone with an academic background in studies on subjective evaluations (which are notoriously hard to nail down), neither the number nor the diversity and balance of subjects is satisfactory. (Of course, Sean knows this very well, but his aim is just to do corporate research directly linked to product development).

    We are talking about taste here. Check out the diversity of restaurants in any bigger city. More or less the same goes for taste in music, sound quality, voicing, equalizing, spatialization, ... , which all depend on age, gender, cultural background, social factors, music genre, listening situation and time of day, with/without video, and other factors.

    This serves to illustrate that the pure notion of having 3 variations of 1 curve, which are supposed to serve all of the above, is not meaningful at all. There is sufficient justification to have not only various wearing styles, colors, comfort features, connectivities, but also voicings. Even McDonalds has much more than 3 items on their menu, and Subway offers quite some variation in seasoning, vegetables, sauces etc., even though some distinguished star cook will insist there is only one "perfect" preparation for chicken fillet and one perfect wine to be served with it.

    When conducting a really scientifically solid study with the required thousands of listeners needed for good statistical significance, which also accuratly represent the demographics of interest, I would expect to find a complex, wide landscape of preferred target curves. If you're lucky, that landscape can be partially parametrized and linked to demographic or other factors of the listeners, perhaps. Other than that, maybe some rather vague features will surface which describe generally favourable or unfavourable curves. Manufacturers might then aim at avoiding the unfavourable features, but if they don't, then those products should fail commercially - or not, which would prove that a sufficient amount of users cares less for sound quality than for other properties of a headphone.
  • This commment is unpublished.
    Dustin · 1 years ago
    Very interesting. Do you know which JBL headphones have been voiced to the harman target with the added bass?
  • This commment is unpublished.
    Brent Butterworth · 1 years ago
    @Mauro No, Sean said the curve is valid only up to 8 kHz. The new "high-res" ear simulators can now do consistent measurements up to 20 kHz or so, but given than the response of human ear canals above 8 kHz varies so much, we don't yet know what >8 kHz measurements really mean.

Latest Comments

Your review helped me to make a decision on getting these headphones.

I'm a super picky ...
Brent Butterworth 11 days ago Focal Arche DAC-Headphone Amplifier
@RoyI'm sorry, I am the furthest thing from an iOS expert. I keep an iPad ...
@Brent ButterworthCan you please tell me how to increase the usb sample bit rate when my ...
Brent Butterworth 12 days ago Technics EAH-AZ70W True Wireless Earphones
@Scott HI'd prefer these if you want noise cancelling, and IMO you need noise cancelling only ...
I'm in the market for my first pair of earbuds and I was deciding between ...
Very, very useful and informative article. Thank you so much!
I found SpeakerCompare to be a decent comparative tool.  As a way to increase sales ...
Michel CZARNY 23 days ago Calyx H Headphones
@jerryThe Thinksound On1 was very good indeed and classy too. Only problem was with the ...
Brent Butterworth 24 days ago The New Standard That Killed the Loudness War
@Soundstagenew2021 Saying the loudness war is over doesn't mean no one's going to create highly compressed ...
Soundstagenew2021 25 days ago The New Standard That Killed the Loudness War
@Soundstagenew2021 Correction discussion is for 2017 album 'Colours', so not new. But if you follow that ...