15 Comments

The Harman curve—the well-known, science-based “target curve” for headphone and earphone frequency response—has been with us for almost a decade. Yet it seems more controversial than ever, and a group of audio enthusiasts who could be called “Harman curve haters” has emerged. I knew this phenomenon existed on some level, but I started to realize how prevalent it has become only after I recently reviewed the Apos Audio Caspian headphones. The Caspians were voiced by a reviewer/consultant named Sandu Vitalie, who describes the Harman curve as sounding “soulless and boring.”

Elsewhere, I’ve seen the Harman curve tarred with adjectives such as “artificial,” “plastic,” and “unrealistic.” I’ve seen it criticized for sounding too hot in the upper midrange—and for sounding too laid back in the upper midrange. In an online poll of 157 audio enthusiasts, 68.8% expressed a negative opinion of the Harman curve.

Sean OliveSean Olive has been in charge of most of the Harman-curve research

As I read the countless online comments about the Harman curve, I’m happy to see some well-reasoned objections, some of which square with my decade of observations based on measurements, my own listening, and comments from panelists in multiple-listener tests I’ve conducted or participated in. But I’m saddened to encounter less-rational criticisms from enthusiasts and reviewers who don’t seem to have read any of the research supporting the Harman curve—and in some cases, who don’t seem to have read much about audio other than product reviews and comments on internet forums.

I thought it might be valuable to examine some of these criticisms in an effort to move the discussion a notch closer to a fair assessment of this research.

First, a quick backgrounder

When the Harman curve headphone research began, around 2010, the science of headphone design dated back about three decades, to an era before computer-based measurements, and it was long overdue for a refresh. Inspired by the emergence of a headphone boom (itself triggered by the invention of the smartphone), scientists at Harman International (parent company of JBL, Harman Kardon, Mark Levinson, Revel, and numerous other audio brands) began conducting blind listening tests to see what type of headphone and earphone sound most people prefer, and if such preferences correlated to measurements of headphone and earphone characteristics.

The research expanded to listening tests involving hundreds of listeners and products, and resulted in 15 (by my count) research papers, reviewed and published by the Audio Engineering Society. As I’ve found in my interviews with headphone designers, other companies, such as PSB and 64 Audio, arrived at similar curves through their own research. And some popular headphones, such as the Sony MDR-7506es, were voiced similarly to the Harman curve many years before the Harman research was published.

Sony

Now let’s consider some of the criticisms of the Harman curve that have emerged. I’ll list them in what seems to me to be the order of validity, with the most rational claims presented first.

1) The Harman curve is fine for average listeners but not for audiophiles.

There’s some validity to this idea, in the sense that audiophiles may want to hear music differently than most listeners do. Companies such as Etymotic, Grado, and HiFiMan won over audiophiles with a focus on products that deliver a bright tonal balance, emphasizing treble and attenuating bass. The extra treble tends to create an enhanced sense of detail and spaciousness. Anyone who regularly hears real instruments in a live setting knows this tonal balance is not natural or accurate, and not faithful to the musicians’ intent, but if some audiophiles enjoy music more this way, that’s fine.

That said, the notion that audio enthusiasts are more sophisticated in their tastes or more accomplished in their listening skills than the listeners employed in the Harman research is way off base. For example, in the company’s main research project on earphone preferences, they used 71 listeners. From that group, 36 had their hearing tested and had passed level eight in the Harman How to Listen training software—and the fact that they completed such a demanding process suggests they had a strong interest in audio and music. On the other hand, I’d guess that few of the Harman curve’s critics have recently had their hearing tested, and probably none could get even half as far through Harman’s training without investing a lot of hours in the effort. Any audio enthusiast who wants to throw shade at Harman’s listening panel should first download the How to Listen software (it’s free), see what level they can get up to, and share a screenshot of their results along with their criticisms of Harman’s methods.

How to Listen

Let’s not forget that the Harman listening tests were done blind, so the listener’s attachment to certain brands couldn’t be a factor—something that no one doing sighted listening tests can plausibly claim. The notion that hundreds of listeners, in carefully controlled tests where the product identity was concealed, would prefer sound that was “artificial” or “plastic” is an extraordinary claim that must be supported with extraordinary evidence to be taken seriously.

2) Everybody hears differently, so an average preference curve can’t work.

This notion is true to some extent because everyone has some natural variation in the physical shape of the ear canal, pinnae, etc., and almost everyone’s hearing has been damaged—from slightly to severely—by exposure to loud sound.

I’ve been conducting multiple-listener tests since the early days of my career, more than 30 years ago. Probably more than any other audio writer working today, I’ve seen first-hand how listener responses to the same product can be substantially different. I’ve also made it a point to work regularly with younger listeners, such as Wirecutter headphone editor Lauren Dragan; she and I had our hearing tested during the same audiologist visit, and she can actually hear up to about 20kHz, while like most men my age, I top out somewhere around 12kHz. Lauren and I do differ in our opinions of headphones and earphones at times, and based on my measurements, I’ve found it’s typically due to things going on in the higher frequencies, where my ears are less sensitive; in particular, a few headphones and earphones that I love, she finds grating because she hears high-frequency peaks that escape my notice.

Interestingly, though, while headphones and earphones that come close to the Harman curve aren’t always the favorites among Lauren, me, and the listening panelists we’ve used in our testing, many Harman curve-ish models have ranked among our favorites, and I can’t recall a single one that listeners sharply criticized.

However, many audio enthusiasts seem to think the claim that “everybody hears differently” means there are radical, unpredictable differences among most people’s hearing, almost as if a gremlin inside everyone’s brain were using some sort of neural EQ to randomly boost some frequencies and cut others. Those enthusiasts often use this notion as an excuse to dismiss any statement about audio that they don’t agree with—but I’ve never seen one of them present data to support this idea, and I’ve never gotten the impression that they’ve done any serious study on the subject of human hearing.

In my work testing hearing aids and personal sound-amplification products (PSAPs) for the technical journal AudioXpress, I’ve had my hearing tested multiple times and interviewed several manufacturers and audiologists on the subject of hearing loss. I’ve learned that while people do sometimes lose hearing sensitivity in unusual ways, most hearing loss is characterized more by degree than by type. Typically, with age, we experience a decreasing ability to hear high frequencies, and often, the development of a so-called “noise notch” somewhere between 3 and 8kHz. People may react differently to some audio products because they have a greater degree of hearing loss, but within a chosen demographic, most people hear fairly similarly unless their hearing has been severely damaged. In fact, some hearing-correction technologies, such as Mimi, offer the option of using the listener’s gender and age to predict the amount of hearing loss.

AES

What’s more, listeners’ preferences are actually far closer than audiophiles might expect. In the 2014 Harman paper “The Influence of Listeners’ Experience, Age, and Culture on Headphone Sound Quality Preferences,” a listening test involving 238 subjects in Canada, China, Germany, and the US, ranging in age from 20 to 70, the researchers found that “Listeners generally preferred the same headphones regardless of their listening experience, age, or country of residence.”

3) It’s just a bunch of scientists telling us what we’re supposed to like.

Anyone who’s read any of the Harman headphone listening test papers understands that this project was an effort not to tell listeners what they should like, but to find out what they like.

Let me state that some headphones and earphones I like fall well outside the Harman curve. Most open-back planar-magnetic models, such as those from Audeze, Dan Clark Audio, HiFiMan, and others, have a much flatter bass response than mandated by the Harman curve, but I still like ’em, and most other reviewers and enthusiasts seem to, too. I’ve recently found a couple of earphones I very much enjoyed—the Campfire Audio Holocenes and the Sennheiser IE 300s—that have a flatter midrange response than the Harman curve, with much less energy around 3kHz.

Sennheiser

I assume Harman began its research as a way to help its headphone brands (AKG, Harman Kardon, and JBL) tune products in a way that would win customers’ approval and increase sales. But the research is equally useful for consumers, most of whom just want to buy headphones or earphones voiced in a way that’s likely to sound good to them, without investing a colossal amount of effort into their selection process. Sure, no matter how a product is voiced, you’ll find someone who doesn’t like it—although we can’t know if that dislike is because of a legitimate objection to the sound or to biases associated with the brand, the design, the technology, or the price.

It’s also important to note, as I did in my 2019 article about the Harman curve, that there are three variants of the curve to allow for varying tastes and hearing characteristics.

Whether headphones or earphones voiced along Harman curve guidelines will be any one listener’s favorite, we can’t say for sure, but once you take the above-noted biases out of the equation, the research shows the majority of listeners will like them. The fact that many headphone enthusiasts object to the Harman curve—in highly biased sighted testing where no controls are in place—probably says more about group dynamics and consumer psychology than it does about audio.

4) The Harman curve sounds “artificial” or “boring” or . . .

Criticism of the Harman curve—or anything else in audio—that relies entirely on non-specific adjectives can be dismissed out of hand. Because it’s reasonable to assume people will lead with their best argument, it’s reasonable to assume the commenter who merely slings derogatory adjectives has nothing more intelligent to say on the subject—which suggests they haven’t amassed much knowledge or experience, and that their comments are unworthy of the reader’s attention.

When audio enthusiasts present only non-specific adjectives in their critiques, they’re making an appeal to authority—presenting themselves as the authority, because they offer no evidence to back up their assertions. Appeals to authority sometimes have merit; if Chris Potter says a saxophone sucks, then maybe it does. But the person presenting themselves as an authority must have considerable and easily demonstrable expertise on the subject, especially if they’re criticizing the work of scientists whose CVs show decades of experience and countless publications.

Chris Potter

Even if a Harman curve hater does make a specific criticism—e.g., the upper mids are too hot or too laid-back—they need to present reasons why we should respect their statement. If they think the Harman curve has too much bass, they have to tell us how they know the Harman curve has too much bass (and specify which version of the Harman curve they think has too much bass). Ideally, what I’d love to see is a well-reasoned technical argument backed with scientific data, such as measurements, controlled listening tests, and citations of existing scientific and technical publications.

In lieu of that, I’d at least like the commenter to list some reasons I should take them seriously. These might include professional experience in audio production or product development; substantial and recent experience in music performance or recording; or documented experience in audio product measurement or audio-related fields, such as acoustics, audiology, or psychology. Anyone who wants to be taken seriously as an audio expert should be able to point to, say, a LinkedIn page that lists their experience and accomplishments, or a Bandcamp page that highlights some of their recording work or performances, or maybe a web page that shows some of the pro-level systems design work they’ve done. Listening casually to a few headphones, gabbing anonymously on audio forums, reading audio product reviews, starting a blog or YouTube channel, or messing around with amateur measurement gadgets won’t cut it.

I hope this article might inspire some of the Harman curve haters to raise their game—confirming their contentions with their own controlled testing, making specific and detailed critiques of where Harman gets it wrong, and publishing and presenting their results so they can be critiqued in public forums. If they really know better than the Harman scientists do, and can make more useful recommendations, I’m confident that the audio industry will be interested. But it’s gonna take a lot more work than slinging a few adjectives.

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

Say something here...
You are a guest ( Sign Up ? )
or post as a guest
People in conversation:
Loading comment... The comment will be refreshed after 00:00.
  • This commment is unpublished.
    Patrick · 7 months ago
    Like the article explained well, there will always be subjective notions to what we as listeners declare to be neutral (as the processing happens within the brain and therefore is your personal interpretation/representation of reality). This also means, that neutral on paper does not mean neutral in the sense of the experience. I've been mastering audio for quite a few years at this point and my target moved from strictly flat to somewhere in the direction of Harman, but with less of a bass boost - yet I usually prefer "some" additional bass compared to neutral nonetheless and tamed top end translates nice to AMTs and planars.
    What usually wonders me the most is the discussion in general. For me the response has to help me in making a track translate to most of your systems, while for you the response is there to make you enjoy the product even more. If you like +30 dB of Sub on your planars and they don't die, well, it's you. I don't think there is a specific "right" or "wrong" in listening enjoyment and variations in target curves can be highly genre dependent. You won't enjoy the same settings on neurofunk and trap that you enjoy on classical, specifically piano :D 
  • This commment is unpublished.
    Rick · 11 months ago
    Some thoughts...
    So first, I'm not familiar with the exact testing conditions used by the Harman surveys, so I'm really not going to comment on their specific target curve. But I'm going throw out some information that might speak to some justification for questioning that one curve fits all...
    1) There were equal loudness curves that were researched originally by Fletcher and Munson at Bell Labs in 1933 that show that a level of perceived equal loudness across the audible spectrum varies by frequency (i.e. phons vs SPL). In general, low frequencies must be increased to match the perceived level at say 1KHz. But what is more interesting for this discussion is that the level of increase required in the bass is greater when the SPL is lower. So an equal loudness curve at say 40 dB at 1kH requires 65 dB (+25 dB) at 50Hz while the curve say at 90 dB at 1kH requires only 105 dB (+15 dB) at 50 Hz. So the point here is, depending on how loud you listen, the louder SPL will sound more bassy. My experience is that different people like to listen at different levels, so that is going to influence how much bass is needed to 'sound right'. (Reference: Master Handbook of Acoustics, Everest and Pohlmann, Ch4  Perception of Sound). (What was the Harman testing like relative to this?)
    2) Ideal speakers measure flat frequency response in an anechoic chamber. The same ideal speakers in room have a rather different response. The high frequencies roll off with distance, there are typically some room modes that cause certain bass frequencies to resonate in standing waves (loud in one place, soft in another), the speaker positions relative to the wall can increase the bass response, porting of speakers also affects how the speakers interact with the walls, a "closed" room may have more sub-harmonic bass level response with unported subs. Well treated rooms with EQ'ed speakers have a more controllable response so you can target them to some response curve, but untreated rooms, which is most of them, are a bit of wildcard. Most consumer speakers are far from ideal with or without EQ. The combination of these may affect the user's preferences for EQ in his listening environment. That not to say that a Harman curve might or might not be a good starting point, but every room is different. There are some sound systems to provided EQ correction and use a mic to measure the response at the listening positioning (e.g Dirac) that will factor this in and probably let you select a room curve as well. EQ will help an untreated room, but the base resonances are still going to cause havoc in an untreated room. (What was the Harman testing room like?).
    3) People will generally have some idea of what a "good" acoustic instrument sounds like. Say for example a "good" violin. But the sound of a violin up close is very different from one in a concert hall. Again the room makes a huge difference. Live recordings occur under all sorts of conditions in all sorts of places. After a performance is recorded, the mixing/mastering will generally attempt to make the recorded source sound 'as good as it can'. This may in some way bring things back to some standard of 'goodness' but there are limitations on how much you can fix a recording. With recordings (like film scores) where they tend to mic each instrument separately, you can probably adjust the recording to taste, but again, it is the mixer's taste.  I think in general, our concept of what an accurate sound system should sound like is based on comparing a playback of a know set of instruments (say a symphony) and comparing it to our recollection of what such an instrument should sound like in a live situation. But it will take more than a few recordings to try to separate the coloring due to the recording from the coloring due to the sound system and the room. (What recordings were used in the Harman testing?)
    4) In cinema (X curve), ATMOS home theater (Dolby), and probably in general, recording studios apply a room curve to their mix rooms that attempts to 'counteract' the room effects of the 'typical' target environment. So for example, the target curve may boost the low end so that the mixer hears more bass, and mixes the base lower because the average target room is going to boost it back up. Similarly, the mix room may roll off the high end so that the mixer mixes the treble higher because the average target room is going to roll off the treble. Different target environments generally need different room curves because - wait for it - the rooms sizes are different. A theater is almost always much much bigger than a home theater.
    4) Back when recordings were mostly on vinyl, there was a limitation on how much bass you can put on a groove and have the needle track it. Because of this, older recordings may tend to have less bass than what gets pumped out thru streaming channels today. Compare say Adele's "Rolling in the Deep" to Beatles' "I Need You". I can say when I was a kid (70's), I'd EQ my parents Hi-Fi console with Bass-10, Mid=0, Treble=6 and it sounded amazing (maybe cause I wanted more bass than was on the source). Some of the more current music tends to be much more bassy, but it's not consistent, especially across genres.
    5) You mention audiophiles and recording studios and say that they want to hear music differently (and perhaps want it to sound bad). I'd say that this set of folks what to hear music accurately. In a recording studio, you mix music in a very controlled environment that has been acoustically treated and EQed to a known state. Then when you mix music in that room, you can hear how it sounds relative to other recordings and mix it so that meets some level of consistency. I'll call this hitting the center spot. If, afterwards, the listener wants to apply a Harman curve, the resulting +10 dB doesn't end up distorting the drivers on whatever grade of gear the user has. And you don't generally have the problem that one song has "too much" bass and another "not enough". The audiophiles are in the camp that they want to hear the music sound the way that it was recorded (as the artist intended). And they will distinguish between recordings they like and do not like based on a myriad of factors that most listeners hear but don't know how to describe in words. (And, yes, some audiophiles are probably tedious and in some cases pychoacoustically deluded - their reviews would collapse in a blind test). This set of folks would not approve of artificially modified bass responses except to the extent it is needed to correct the room response. As a target curve (i.e. measured room response), they would pick a flat low end response - not a +10dB curve. 

  • This commment is unpublished.
    a person · 1 years ago
    Harman curve has too much bass.  It's right there on the graph.  The Harman (IEM) curve is a preference curve based on an average of the preferences of a handful of Harman employees.  It's an average.  The actual preferences varied from subject to subject.  People, on average, prefer a bass boost and a treble boost.  Not everyone does.  A different subset of people might prefer a mid boost, or a perfectly flat response.  You even mentioned that audiophiles prefer a flat response with elevated treble. 

    It's as if you asked every American what a beautiful woman looked like and then they averaged all the results and came up with a 5'7" strawberry blonde with an hourglass figure.  Then every actress and model on TV targeted that look.  If you complained about it because you were not attracted to that, you are scientifically wrong.  If you disagree, then your opinion doesn't matter unless you are an expert in hot women, and you better provide evidence of your experience with hot women.
    • This commment is unpublished.
      Brent Butterworth · 1 years ago
      It says right in the article that the IEM curve was derived from tests of 71 listeners, which I would consider more than a handful. Harman has used its own employees, but has also done numerous studies with a wide range of listeners around the world. There are preference differences among people, but research shows they are seldom dramatic. And if you want us to respect your contention that Harman curve has "too much bass," explain to us why we should. How do you know what too much bass is? Do you have anything in your CV that should make us respect your opinion more than years of research by some of the industry's most experienced scientists? Lay it on us, please.
      • This commment is unpublished.
        Doug Schneider · 1 years ago
        This is a question I hope Brent can answer -- is it true they are "revising" the IEM curve? I read that somewhere, but also wonder what that means, exactly.
      • This commment is unpublished.
        common sense · 7 months ago
        "Too much bass" is too much bass if you think that the bass is too loud. Dr. Olive and his research explicitly states that the bass can be tuned to preference and varies a lot between different listeners, so I don't know why you and this article uses bass as an example here.
    • This commment is unpublished.
      Doug Schneider · 1 years ago
      Brent can answer more, but the curve is simply a "preference" curve -- not an exact curve. They've done the same with loudspeakers. Furthermore, with loudspeakers they have also done studies regarding the amount of variance in terms of preferences -- and it is a lot. So you're right, it might be too much bass -- but it might also be the right bass or too little bass.
  • This commment is unpublished.
    thermglass · 1 years ago
    This post is very informative and helped me lot for getting a understand of harman curve sound and find best Vacuum Insulated Glass which is sound proof for practicing instrument in home. https://thermglass.com/
  • This commment is unpublished.
    Dave Angelo Obico · 2 years ago
    Heres my take. The etymotic target for iems sounds better for me with a boost bout 20to 35hz region. While the harman is not. However. For speakers? Harman takes the cake for me.
  • This commment is unpublished.
    Daniel Schröder · 2 years ago
    My personally heard frequency response of a flat signal through a Harman-tuned inear would sound NOT FLAT, which for me equals to NOT IDEAL. Crinacle's IEF Neutral hits it a lot better, but not perfect either.

    I do not have a perfectly Harman tuned IEM. I try to combine two approaches:
    a) deviations between MY personal perceived neutral (no volume changes) to what I hear over my inears while playing back a neutral-volume sine sweep
    b) the deviance of the accurately measured IEM (old Aether R) from a target frequency response curve, like e.g. Harman.

    I can extrapolate (purely for myself!!!) that the Harman target, while sporting a little too much bass overall (difficult to say how much, I'd say 3dB), it lacks a sub bass rise below 30Hz (~5dB maybe?), it lacks a little upper bass / low mids between 300--1000Hz, the pinna gain is slightly exaggerated (by maybe 3dB?), the high mid / lower treble section is okay / my reference point and its upper treble drop over 12kHz is just abysmal. I hear well until 16kHz and I can still make out 18kHz (traffic as a pedestrian (when I do not wear IEMs) is a nightmare of brake sounds for me), but the Harman upper treble still sounds far too recessed, I can't even make out by how many dB.

    This of course will vary for every person. :)
  • This commment is unpublished.
    GoneToPlaid · 2 years ago
    Personally, I prefer the IEF Neutral Target curve over any of the Harman curves.
  • This commment is unpublished.
    Douglas Barnes · 2 years ago
    Thanks for hosting a rare civil discussion of a topic that seems to generate heat more often than light. While not a Harman Hater, I confess that I am skeptical. More broadly, I question the idea that available measurements (e.g. https://www.audiosciencereview.com/forum/index.php) can really tell us how a set of headphones will sound. At best, they help flag certain flaws (distortion; idiosyncratic voicing), and they are unable to capture other elements that may be equally important, if not more so.

    In my experience, whether a headphone sounds "right" becomes apparent from listening to a well-engineered recording of a string quartet, classical or jazz piano, or classical vocal ensemble. I'm thinking of labels like BIS, Gimmell, and MDG, which record artists in natural acoustics with (relatively) simple mic setups. 

    My first venture into hi-fi headphones was a Grado set I bought 30 years ago. I took a few CDs --rock, jazz, and classical-- to the local hi-fi salon, endured the stifling attitude, and compared about five headsets, including Beyer, Stax and a couple others long since forgotten. The Grados were they only ones that didn't sound somehow "off" with a Haydn string quartet. What surprised me was that the differences between sets were not subtle: they were apparent ten or twenty seconds in.

    I think this is not about the overall spectral balance, as is captured by the Harman Curve, but truthfulness --accuracy, if you're in that tribe. Grados (those I've heard) are indeed bass-light and boosted in the upper midrange. But they have a house sound that is consistently truthful to string instruments, voice, and piano. I once read that Joe Grado was an avid opera fan. 

    Headphones I've heard over the years have failed the quartet test as often as not. I strongly suspect this is down to the overtone series of string instruments, which become very recognizable once one has heard them a few times in a concert hall or parlor. When the cello insists on whining like a bumblebee, it's probably a subpar recording (e.g. 1960s and '70s Melodiya), or something is wrong with the voicing of the headphones. The same holds for loudspeakers, for what it's worth.

    How does this relate to measurements? The plots for headphones are often quite ragged from 4-5 kHz upward, with swings of 6 dB or more from octave to octave. If not properly managed, they can make a million-dollar Strad sound like a kid's Suzuki starter model, as a commenter on MassDrop (a Russian, natch) astutely put it. 

    They may be out there, but I've yet to find headphone tests focusing on distortion and the Harman Curve that are also attentive to what real instruments sound like through the sets under test. As a consequence, I rely on in-person auditioning first and foremost. If that isn't possible, I seek out a reviewer (e.g. Keith Howard when he was at HiFi News) who has listened with music I know well.
  • This commment is unpublished.
    Napier Lopez · 2 years ago
    Great piece Brent! Really captured a lot of my thoughts on the topic with a clarity of language I usually can't muster when trying to debate these things. At the end of the day, I've very rarely seen Harman Curve haters present reasonable counter-arguments or cite valuable opposing research. Heck, it'd be fun to see science to the contrary, but I haven't seen it.

    And like you, I've also liked headphones that aren't the most 'Harman-y' in the world. Sometimes there are headphones that don't sound super neutral, but have impressive spatial presentation, for example.  But overall, it's very rare that I dislike a harman-curved tune headphones; usually, it's because of something outside of the actual tonality, like a really high noise floor in some wireless headphones or obvious distortion. Occasionally there are headphones that get close to the target but have a stray peak I find unbearable, and I don't begrudge someone for loving a headphone that measured totally weird. But that doesn't change the trend of harmanesque headphones usually being better, and therefore usually being safer buys than ones that aren't.
  • This commment is unpublished.
    z mitchell · 2 years ago
    The Harman curve isn't a measure of accuracy, but a consumer preference curve.
    • This commment is unpublished.
      Mauro · 2 years ago
      Actually I think it is a mix of the two: preference on bass and treble on top of the translation of a good speaker in a room to a headphone 
    • This commment is unpublished.
      zeeman · 2 years ago
      2 downvotes for simply pointing out that the Harman Curve was conceived due to research into consumer preference in sound, and not accuracy in sound?  Where does it ever say it had anything to do with accuracy?
    • This commment is unpublished.
      Napier Lopez · 2 years ago

    • This commment is unpublished.
      Napier Lopez · 2 years ago
      You are correct, that the Harman curve is a measure of consumer preference, but that does not preclude it from being a measure of accuracy as well.

      The thing is: as far as human perception is concerned, preference is accuracy. This has been demonstrated time and time again with speakers, going back through decades of research and numerous studies. The speakers most preferred in double blind tests are nearly universally described as the most accurate speakers, even when these are addressed in separate remarks. It also just so happens that the most accurate/preferred speakers tend to be those with a flattish (anechoic) frequency response and smooth directivity -- the ones that most directly translate the recorded signal into sound waves.

      Intuitively, this isn't surprising. We want music to sound like live music. We want voices to sound like real voices -- our most common reference for "live sound.". Where differences among groups exist tend to be in the bass because some people prefer the physical tangibility of heavy bass, and treble mainly because of hearing loss. But these differences are usually not dramatic and on the whole, people prefer flattish speakers with smooth directivity.

      It is not big a leap that headphones should be similar -- that preference correlates tightly with the perception of accuracy. More concretely though, the Harman curve looks a whole lot like the frequency response of a good speaker in a good room, if measured at the ear canal after modification by typical ear geometry. And because music is mixed for speakers, it makes sense that the best headphones are typically the ones that most approximate the tonality of speakers.

      There are, of course, ways that the Harman target could be improved, especially for individual listeners. But as a rough overall target for neutrality and accuracy, it really makes a lot of sense.

      • This commment is unpublished.
        Brent Butterworth · 2 years ago
        Great explanation. It's hard to get audio enthusiasts past this idea that sound needs to be altered for consumer consumption, when so many manufacturers and publications push the idea. 
  • This commment is unpublished.
    Blaine LaCross · 2 years ago
    There are many valid complaints to raise about the Harman research - from the low resolution of the smoothed in-room "baseline" used, the essentially arbitrary selection of the filters end users were allowed to adjust in Listener Preferences for In-Room Loudspeaker and Headphone Target Responses, the relatively poor engagement with prior work on the diffuse field target, and more.

    All of these complaints, however, really have to either reconcile themselves with, or convincingly challenge Olive, Welti, & Knonsaripour's papers documenting robust prediction of listener preferences using a statistical model whose input is, itself, the headphone's response and the Harman target (well, strictly, a few variables based on the relationship of the headphone's response and the target).

    There are a lot of ways I'd like to refine the Harman target, and I had the privilege to discuss them a bit with Dr. Olive on a podcast recently, where he seemed amenable to some feedback, but the elephant in the room is that no matter what sticking points we have with the details, when listeners are blinded to anything other than sound quality, headphones falling within a certain tolerance of the Harman target win over those which are further away.

    I don't think you're going to see much engagement from the big critics of the Harman work, however - as you note, they generally haven't read any of the work, and when they do try to make technical criticisms, this typically becomes obvious. I'm not sure it's really possible to have engaged with the body of the Harman headphone work without being, at least to some extent, a fan, simply because it produced a lot of good science which can be carried forward into future work, even for those of us whose preferences fall towards the edges of the distribution.
    • This commment is unpublished.
      Brent Butterworth · 2 years ago
      Good comment, thanks! Yes, it's easy to criticize scientific studies, because they all are done with limited time and resources. I always say the perfect test exists only in the minds of people who don't actually conduct tests. Until I see a documented study that suggests Harman curve doesn't predict user preference, I'll continue to consider it a useful guideline.
  • This commment is unpublished.
    Gary · 2 years ago
    @Brent Butterworth what's the highest level you can pass on all tests of Harman's How to Listen?
    • This commment is unpublished.
      Brent Butterworth · 2 years ago
      I can't remember, I did it many years ago when they first released the beta. Planning to do it again after the holidays. I do remember than it took me a long time to get up to a decent level, and I definitely didn't get to 8.
  • This commment is unpublished.
    Kevin Voecks · 2 years ago
    Well done Brent; Bravo!
  • This commment is unpublished.
    Jon · 2 years ago
    Hi Mr. Butterworth,

    Thank you for your well thought out and scientific articles. You are one of only a few audio writers I trust online. I have been meaning to ask you something about the Harman curve, and this article seems as good a place as any to do so. But first, a little bit about me, as you insist at the end of the article.

    I am a dirt cheap listener. My currently most used audio gear is a $13 TWS JLab audio air pop Go. But I have been playing guitar as a hobby (and a little bit of bass and drums) since middle school (over 20 years). I was my church's live audio guy for a few years, and when I began that role I read a couple of books on the science of audio engineering. I have taken the Harman hearing test, and I think I got to about level 6. Though my gear has always been dirt cheap, I know what music is supposed to sound like and I can identify what frequency ranges the problems are in at an amateur level. I don't think I can hear anything above 15k.

    My experience with the Harman curve is through EQ profiles provided by the GitHub AutoEQ project. I have read whatever Harman research is made available for free online.

    My question about the Harman curve is this. In my experience, when I EQ headphones to the Harman curve, the range between 1k and 2.5 k seems 2-3 db too strong, introducing coloration. It feels to me like looking at the world through orange sunglasses. May be a better way to say it is to say that it sounds a little bit like radio broadcast rather than in-person sound, though not to the same degree. I think I find this to be the case pretty consistently. So I regularly EQ that range down by about 3 db after tuning to the Harman curve.

    I thought it was just me, but the last few headphones you reviewed where you said the headphones did not have any coloration and were clear and detailed without boosting the treble had deviated from the Harman curve in just this way, where the 1-2.5 k range was underemphasized relative to the Harman curve. I think the latest drop Sennheiser collaboration HD 8xx were also tuned with this range cut (though apparently the audiophile community hated it).

    It got me thinking. Is it possible that the way that the Harman curve was derived, taking as a starting point what flat speakers in a listening room sounds like at the ear drums, then bass added to adjust for room effects, was problematic? Is it possible that the ears are more sensitive to ear gain when music comes from so close?

    But then again, I also regularly feel that music coming from speakers in public places also have this coloration.

    The possible conclusions I can draw from these observations are as follows: 1. Many mastering engineers don't have flat speakers or good listening rooms (I doubt this); 2. Many mixing engineers like this mid range coloration; 3. the Harman target may have too much energy in the 1-2.5k range; 4. Tuning to Harman with EQ has problems (though from my reading, Harman uses this method for their double-blind tests); 5. My ears are useless.

    I was wondering with the Harman method of deriving their curve, if instead of giving the listeners the option of boosting treble or bass they had the option of cutting that midrange they would have opted for that.

    Again, all this was prompted by your description of headphones that measured below the Harman curve in that range as natural, without coloration, clear and detailed without artificially boosting the treble. May be the Harman curve is problematic in that area and we have all just been trying to EQ it out by adding bass and treble instead of cutting the mid? Your thoughts would be much appreciated.
    • This commment is unpublished.
      Brent Butterworth · 2 years ago
      Thanks for the thoughtful and detailed comment! The Harman curve offers three variants for user preference, which I discussed in my article here: https://www.soundstagesolo.com/index.php/features/217-where-are-we-at-with-the-harman-curve

      So I don't know which one you're referring to -- and neither do any of us, I would guess, because TMK no one, not even Harman, promotes which version of the curve they're using in the marketing materials for their headphones. (I really need to do a better job of figuring out how to present all this in my reviews, but it's tough, because Harman to date has provided me only with one OE and one IE curve, both created using non-standard test gear.) It's possible that the Harman curve variant with more bass would balance out the excess upper mid/lower treble energy you're hearing. Or it's possible you fall within the X percent of people who don't prefer Harman curve. 

      The in-room target was just a starting place for them -- one that happened to work -- and they're showing overwhelming user preference for that, so I don't see it as a problem. 
      • This commment is unpublished.
        Jon · 2 years ago
        Thanks for the reply. Your reply prompted a few more thoughts, so if you would be so kind as to address a few more questions, that would be really appreciated.

        1. Where I said Harman curve in my first comment, I was actually referring to the target curve used by the AutoEQ project on GitHub, which you can read about here: https://github.com/jaakkopasanen/AutoEq

        I have read their methodology, but I am still uncertain as to how exactly they adapt their target curves based on the source of the measurements and the various measurement rigs used.

        Are you familiar with this project at all? I think I read you say something about it in an interview you did on Soundstage Access. What are your thoughts on EQing headphones in general? Do you try to EQ headphones that you are not reviewing? You may have written about this topic before, so if you have please point me to where I can read your thoughts.

        2. I have read those articles about people who prefer a little bit more bass or a little less bass than Harman. When I first started trying to EQ to Harman using the AutoEQ project, I thought I was just a basshead. And then I felt I also needed a little bit more treble, so I thought I was the typical consumer who enjoys v-shaped or u-shaped sound signatures. But when I listened carefully I could still hear a slight coloration in the upper mids. I believe I am hearing the overtones more than I am supposed to relative to the fundamental frequencies. It was only when I EQ'ed the upper mids down that I felt it sounded natural. So I tried again, starting with EQing the upper mids, and I realized I did not need to boost the bass much at all, though I still enjoy a little extra treble boost above 10k (I may just be getting old). This is why I asked about whether if listeners in the Harman research had the option to dial in the upper mids first they may be more reluctant to add bass.

        3. I realize you are trying to be as objective as possible when speaking about this stuff, so you have not said anything about your personal preference. But if you are comfortable sharing, in your opinion, does your personal preference match the Harman curve? As mentioned in my first comment, in some of your recent and most favorable reviews, you seem to enjoy a similar deviation from Harman as I described. I am referring specifically to your review of the Shure Aonic 215 Gen 2, 64 Audio Duo and U6t, Apos Audio Caspian, and there may have been a few others. In all these reviews, you comment on how balanced, natural, clear and detailed they are without artificially boosting the treble. All of these headphones are underemphasized around 1-2.5khz relative to Harman. Would it be an unfair caricature to say that you prefer a little bit less energy in the 1-2.5khz range than Harman?

        Thanks again for engaging me in this conversation.
        • This commment is unpublished.
          Brent Butterworth · 2 years ago
          1) Yes, I've looked at the AutoEQ project a little bit, but I can't say I've really engaged with it. I do think EQ'ing headphones can be a good idea, and I've done it with many models I haven't reviewed. If I do it in a review, it's usually because the headphones are not far off and could benefit from just one (preferably) or two small tweaks of an EQ slider.

          One thing I have to add, though, is that I've found I get used to the sound of certain products over time. I used to think my Jabra 65ts sounded way too dull (I bought them because they're super-waterproof and secure for use during workouts), and I went into Jabra's app and dialed in some treble boost, but over the last 3 years, I eventually started to get used to the native sound of them. Still not my favorite but they're OK.

          2) I just don't know. From doing panel tests, I've learned that it's really difficult to predict how any headphone is going to suit one person's preferences. Although Lauren Dragan at Wirecutter has, over years of working with me, gotten very good at guessing what I'll like. Part of that is probably because I give her detailed critiques that are based on experience listening and measuring other headphones.

          3) Yes, I think you are probably right here. That squares with what I've been thinking. 


          • This commment is unpublished.
            The big pleh · 2 years ago
            Have you seen some of the newer DSP tech used in some understated headphones like Dirac's virtuo used in Cleer Audio's Alpha? Yamaha has something called the 3D Sound Field which I assume means they tapped into their YPAO research to develop their own DSP Spatializer. I think theres a bit of tech coming w/ DSP correction that will overlap w/ the Home theater market. 
            • This commment is unpublished.
              Brent Butterworth · 2 years ago
              I haven't tried the Cleer Audio Alpha yet, but I have tried the Yamaha, which didn't impress me. I'm hoping that as the DSP inside headphones/earphones gets more powerful, we'll get some more effective and appealing surround simulation and HRTF processing. I know it's possible, but so far the only really impressive processing I've heard depended on the use of insert mics to measure the user's HRTF.
          • This commment is unpublished.
            Jon · 2 years ago
            Thanks again for your reply. When I was in charge of my church's live audio, making sure that I new what was neutral was very important for me, but since I'm not in that role any more I just want to enjoy my music. But I also don't want to be enjoying something that may be wildly inaccurate. I do want to hear music as the artists intended it before I apply a little EQ to enjoy it. In that respect, it's comforting to know that my preference may not be very far off from yours, a musician and respected headphone reviewer.
            • This commment is unpublished.
              Brent Butterworth · 2 years ago
              I feel exactly the same way. IMO, job #1 for reviewers should be to weed out the stuff that's wildly inaccurate/unnatural. Sadly, many reviews embrace these kinds of products -- which I think they wouldn't if they just went to Guitar Center, grabbed a drumstick and tapped on a few cymbals.
              • This commment is unpublished.
                Jon · 2 years ago
                Thank you. This is why I think it matters to me that you are a musician and you know what you hear when you play your instrument and when your band plays. There was a season in my life when I was obsessed with trying to get my acoustic guitar plugged in to sound exactly like it sounds mic'ed (as I'm sure you know, the tech for doing this properly is still undergoing major development because it is still far from perfect). When you obsess over things like that you pay a lot of attention to what sounds natural. That's the sort of thing people who don't play music just won't understand.
  • This commment is unpublished.
    Mauro · 2 years ago
    I stand by Harman curve and no one of us plain readers will be able to recreate a statistically meaningful study on preference curves, let alone something better. It will likely become an industry standard. 
    But 
    An A/B EQ-based blind test can be conducted alone in our little dark rooms 😄, a Mimi hearing test can be done downloading an app, the same applies to the Harman testing suite which is free. 

    I think this leaves room for us readers to express some specific and informed criticism. For example I commented that that the AKG K 371 has some shrillness and it is really in your head/face. I’d like to read a future publication on the emerging immersive audio processors for headphones with HRTFs to mimic a speaker OUT of your head. I guess they might complement and progress from Harman Target curve findings. 

    Nice article.. and good luck with all the spicy answers you are gonna receive 😇
    • This commment is unpublished.
      Joseph Yeung · 2 years ago
      Oh I have no doubt that audio processors can improve greatly on the Harman curve, because they have many more tools at their disposal than simply tweaking the frequency response.

Latest Comments

Carter Mills 1 days ago DALI IO-4 Bluetooth Headphones
They don’t have noise canceling, they’re super comfortable, and I appreciate the reliable push-button controls. ...
Amarres de amor Houston TX 1 days ago Ausounds AU-Flex ANC Bluetooth Earphones
 Amarres de amor Houston TX are love rituals used to make relationships stronger. People use ...
Jimmy John 8 days ago DALI IO-4 Bluetooth Headphones
They don’t have noise canceling, they’re super comfortable, and I appreciate the reliable push-button controls. ...
Kirkland Signature 8 days ago Ausounds AU-Flex ANC Bluetooth Earphones
Kirkland Signature is a brand you can find at Costco. They sell many things like ...
Kirkland Signature 8 days ago Technics EAH-AZ70W True Wireless Earphones
<a href="https://kirklandsignatures.us/">Kirkland Signature</a> is a brand you can find at Costco. They sell many things ...
io games 12 days ago DALI IO-4 Bluetooth Headphones
If you're considering purchasing Ozempic, a medication used for managing type 2 diabetes, you might ...
Angie 25 days ago Yamaha YH-5000SE Headphones
I've been using the SANWEAR-GT gaming earbuds, and they’ve completely changed my gaming experience! I ...
headphonejack 1 months ago Is Accuracy the Only Option?
Is accuracy measurable? What do you consider the most accurate headphones? Or maybe top 5?
Doug Schneider 1 months ago Is Accuracy the Only Option?
It's an interesting point that's made. The only thing I'll say contrary is that with ...
Doug Schneider 2 months ago PSB M4U 9 Headphones
@steve pageThis problem is an interesting one. If you see this post, let us know what ...