Harman preference curve for headphones - am I the only one that doesn't like this curve?

[Blake]

I'm baffled that the Harman curve is so controversial. For me, its use as a baseline has revolutionized my approach to listening to headphones. While I generally like the curve, I find it a shade bright and tend to adjust the higher frequencies downward. This has resulted in the most pleasing sound I've ever gotten. I couldn't have done that without that reference point.

Pre-Harman was like adjusting the color on my TV set in the 1980s with no instructions or reference. "Okay, flesh tones are maybe a bit green still, but I think that's as close as I'm getting." I would try to get a rough idea of what headphones sound like from reviews, maybe seeing frequency response graphs without fully understanding how that would translate to the sound in my ears. Maybe I would find a subjective opinion on how it sounds compared to a common headphone, typically the HD 600 or 650/6XX, which was a looser form of reference. I would typically listen to headphones without EQ, but if I did apply any, it would be more or less blindly, just trying to get a feel for what adjustments were agreeable, rarely hitting on something I was happy with. ("Flesh tones look right, but now the sky is purple.")

Post-Harman is like having a modern TV with color calibration equipment. I can get the color as close as possible to an agreed-upon reference, and then adjust further if I still prefer, say, a brighter image with cooler tones. Generally I do that by loading oratory1990's settings and adjusting from there. But now I'm adjusting from a known reference (more or less), making my adjustments much more predictable to me.

So, while I don't exactly prefer the Harman curve, I am very grateful for it. It's a baseline, not a requirement.

 

[Doodski]

I'm baffled that the Harman curve is so controversial.

The beauty of the Harmon Curve is that if the curve is there then the EQ/PEQ can accommodate most any EQ/PEQ adjustment the user desires. If the curve is inaccurately all over the place and is inaccurate then a EQ/PEQ cannot compensate due to limitations in the EQ/PEQ parameters.

 

[jae]

I assure you the blue graph doesn't reflect how the Ether C Flow sounds vs. the Stealth. Where is it even from?

Your comments about ear gain are appreciated and educational. But to equate the two headphones in question from 500 Hz down based on just this, and not experience that draws on how differences in attack/decay, group delay, diaphragm deveation from a piston, ringing, reflections and resonances from infront of and behind the diaphram, sound into and thru the pad, the angles all of that make entering the ear canal, much more, exemplifies the problem with comparing just FR graphs. And I'm just as guilty as anyone else in doing that.

The data shown on the graph are from Crinacle's GRAS 43AG-7 measurements. Assuming they were taken properly, which I have no reason to believe they weren't, they should be compatible with Amir's measurements here (GRAS 45CA-10), the original Harman research, the original compensated Harman curve. I also think this is a compatible fixture to what Dan used to design the headphone to this tuning, but don't quote me on that. You can make the argument that how these headphones sound to you or the frequency response at your ear drum will be different to a test fixture, however it does not take away from the obvious differences from one model to another on the same test fixture.

The measurements I provided are quite high resolution, 1/48 octave, so these lines are made with ~500 data points for each trial, multiple trials averaged, with half of those being data up to 500 Hz. Headphones in general are minimum phase devices, at least in most normal conditions, these well-measuring headphones even more so- if you understand the concept of how a frequency response is generated, transform functions and time domain, you will understand that practically all of these "differences in attack/decay, group delay, ringing, resonances" and all sorts of other issues like nulls/phase cancellations etc. will all show up and manifest itself in a simple frequency response plot, often proportionally to its influence in the time domain if the resolution is acceptably high. If you want to comment about group delay, I don't believe there is any difference- the Stealth and even other years old DCA headphones also all have pretty much non-existent and super clean group delay (look at the Ether CX, A-RT, and Stealth reviews on here). If you think this is in any way audible, I would suggest reading Toole's books or some other psychoacoustic texts discussing the audibility of such phenomena.

All this aside, both headphones in the graph I provided were measured on the same fixture, by the same person, in the same conditions, (at what I assume are totally different times), and averaging multiple measurements taken. So the fact they are similar is telling at least something. And of course, the drivers/headphone design are of course both Dan Clark's, so knowing that info I don't really have any reason to believe they would not be similar if I had to take a bet. Ether C flow was their flagship closed headphone before the Stealth, so it makes complete sense to me that he thought to himself he had something great with the bass on the C flow and great feedback, and that the improvements to the new model (stealth) came mostly from the fit and comfort, distortion, "refining" the upper frequencies or "fixing" the ear gain region, which many of his previous headphones had "issues" with. I don't doubt the drivers used in both are probably very similar. I'm using a lot of "quotes" here, because of course everyone has a preference, and maybe these past "deficiencies" even became the Dan Clark "signature sound". But, these shortcomings were also the brand's biggest criticism and what most people complained about even if they liked other facets of the headphone. It's clear he was out to design a reference headphone, and nothing is more of a reference than Harman's research. I've tried virtually every Dan Clark open headphone (except the RT open), and thought holistically they were good products and I was so close to buying them so many times, but each had obvious problems to me which lead me to not purchase them and wait for a refinement. As much as I want the Stealth, I have not bought it yet because I don't like wearing closed headphones for long periods of time.

 

[jae]

I'm baffled that the Harman curve is so controversial. For me, its use as a baseline has revolutionized my approach to listening to headphones. While I generally like the curve, I find it a shade bright and tend to adjust the higher frequencies downward. This has resulted in the most pleasing sound I've ever gotten. I couldn't have done that without that reference point.

Pre-Harman was like adjusting the color on my TV set in the 1980s with no instructions or reference. "Okay, flesh tones are maybe a bit green still, but I think that's as close as I'm getting." I would try to get a rough idea of what headphones sound like from reviews, maybe seeing frequency response graphs without fully understanding how that would translate to the sound in my ears. Maybe I would find a subjective opinion on how it sounds compared to a common headphone, typically the HD 600 or 650/6XX, which was a looser form of reference. I would typically listen to headphones without EQ, but if I did apply any, it would be more or less blindly, just trying to get a feel for what adjustments were agreeable, rarely hitting on something I was happy with. ("Flesh tones look right, but now the sky is purple.")

Post-Harman is like having a modern TV with color calibration equipment. I can get the color as close as possible to an agreed-upon reference, and then adjust further if I still prefer, say, a brighter image with cooler tones. Generally I do that by loading oratory1990's settings and adjusting from there. But now I'm adjusting from a known reference (more or less), making my adjustments much more predictable to me.

So, while I don't exactly prefer the Harman curve, I am very grateful for it. It's a baseline, not a requirement.

I like your analogy, I am an arcade/CRT enthusiast and it hit right at home for me. I've in the past likened it to finding the right colour temperature/white balance for a display while knowing the video was produced and mastered for a specific standard that has some sort of validity through ubiquity. Harman is representative of this ubiquity, even if it is not an official standard is can be interpreted as one by virtue that significant majority prefers and enjoys it, and it is the best we have. Regarding simplifying and revolutionizing the way I listen with headphones, I couldn't agree more and I came to a similar conclusion ages ago, and so has virtually everyone I've explained this to or tried to help with tuning their headphones. Even those that dont adhere to it completely takes some facet of the approach, whether is it the location or amplitude of the bass shelf, the amplitude and breadth of the ear gain, or the relative ratios/relationships between these critical areas.

 

[markanini]

I'm baffled that the Harman curve is so controversial.

It's certainly not controversial to people that understand it's purpose. Most of the "debate" boils down to people honestly trying to clear up misconceptions and critics refusing to listen. Even Sean Olive himself responded to a critic at one time and he still acted like he knew better.

Some may say this is just people with another opinion. Well isn't that great. Simply by a few people using grade school social tactics non-issues turn into controversies. I think there's a discussion to be had about how people like that are in effect given a sanctuary from where they're able to set the narrative.

I know on forums that are ad-driven total engagement matters more so TOS tends to be followed rather shallowly. For ASRs purpose, of promoting discussion about objectivity in audio, you would think lower engagement of a higher quality would be more valued OTOH.

 

[Feelas]

It's certainly not controversial to people that understand it's purpose. Most of the "debate" boils down to people honestly trying to clear up misconceptions and critics refusing to listen. Even Sean Olive himself responded to a critic at one time and he still acted like he knew better.

Some may say this is just people with another opinion. Well isn't that great. Simply by a few people using grade school social tactics non-issues turn into controversies. I think there's a discussion to be had about how people like that are in effect given a sanctuary from where they're able to set the narrative.

I know on forums that are ad-driven total engagement matters more so TOS tends to be followed rather shallowly. For ASRs purpose, of promoting discussion about objectivity in audio, you would think lower engagement of a higher quality would be more valued OTOH.

Yet it must be said that in all the amount of sound-related drivel, marketing and bullshit-tactics (sorry mods, but I can't help calling it that way), it is impossible that people won't be skeptical towards seemingly "yet another sales tactic" devised by Harman. Without any measurable effort it is impossible to differ between the two.

I think that one singular thing that is kept apart and might speak to people is that (going away from headphones for a sec) Floyd Toole has committed testing, which showed that in given that blind testing is properly done, people seem to drift towards more neutral sounding gear EVEN when you swap the anchors. The relative ranking of quality in blind testing stays the same, even if gear in question starts differing less and less, the ratings seem to stay in the same relation. One must grasp that subjectivity might be a mirror of objective truth, not merely "that subjective is elsewhere, but surely not where the measurements are".

 

[192kbps]

Our approach was to simulate the smoothed response of the headphone above 12kHz but not try to simulate or EQ high and medium Q resonances because there are HF errors from variations in the positioning of the headphone on the coupler and the listener's ears. Add to that the accuracy of all IEC couplers are questionable above 10kHz in terms of simulating human ear canals and the acoustic impedance.

We tried listening to headphones with different filtering above 12 kHz, and frankly we could not hear huge differences except slight ones with pink noise. Maybe we are too old but "we" included Omid who was in his late 30's.

So, I don't know the answer to your question except for many people these details above 12 kHz are probably not audible or as important as below.

For that reason our predictive model only considers deviations from the harman target up to 10kHz.

The Harman Target Curve was based on a measurement at the primary listening seat in our listening room using Revel F208 and a JBL M2 calibrated to the Harman Speaker Target. We measured it with our ear simulated mounted in a head/torso. We did a spatial average +- 30 degrees by rotating the head.

"We measured it with our ear simulated mounted in a head/torso. "

GRAS 43AG?

 

[skris88]

Not sure that I understand what you're suggesting here, Zolalll.

Zolall is absolutely correct.

What he's saying is that EQ correction curves are affected by the playback level.

If it's designed to be played at an average loudness (C weighted, Fast) of 83dB (SMPTE standard), it'll sound balanced both in the bass and treble.

But if you're playing at 88dB, it'll be too bassy.

Or if you're playing it at 78dB, thin and bass weak.

Remember, 83dB is Average loudness.

Instantaneous Peaks can be as loud as 110 dB - and still be balanced (ie. not too boomy, thin, shrill or dull).

I have a theory/personal view that people tend to listen to modern popular recordings at 78dB - since they're all compressed as to their dynamic range. This will explain the 6dB bass shelf from 105dB down added into the Harman headphone curve. Since I tend to listen at 83dB, I found it bassy.

But, easy solution from http://autoeq.app website, where I could remove the bass shelf and download a new target - which I am using with my various headphones ("...my hobby is collecting headphones"!)

Lastly, to the person who added it, love that link to Dr Sean Olives July 2023 post, thanks!

I am researching to confirm if my theory above is right...

Enjoy listening, everyone!

 

[Blake]

I am researching to confirm if my theory above is right...

I don't think you have to. It seems like you're just describing the Equal-Loudness Contour. The observation fits. To be perceived "equally loud" across the audible spectrum, a softer signal (dB SPL) would need to have the bass boosted compared to the mid/higher frequencies. As volume increases, bass doesn't need to be boosted as much to be perceived as "equally loud" compared to the other frequencies. That's exactly as you describe: If the whole spectrum was left constant as volume increases, the perception should be that the bass becomes relatively louder/more emphasized.

This to me still reinforces the value of Harman as a baseline, though. All I need (and all I personally use, as a general rule) are bass and treble shelves to make minor tweaks to my taste once I've applied a Harman target EQ. If I started listening at a different volume than I usually do, I'm sure those tweaks would be somewhat different.

 

[markanini]

Does it need to said? You don't have to question Harman as a baseline to have a valid reason for diverging from Harman for personal use.

 

[Feelas]

Zolall is absolutely correct.

What he's saying is that EQ correction curves are affected by the playback level.

If it's designed to be played at an average loudness (C weighted, Fast) of 83dB (SMPTE standard), it'll sound balanced both in the bass and treble.

But if you're playing at 88dB, it'll be too bassy.

Or if you're playing it at 78dB, thin and bass weak.

Remember, 83dB is Average loudness.

Instantaneous Peaks can be as loud as 110 dB - and still be balanced (ie. not too boomy, thin, shrill or dull).

I have a theory/personal view that people tend to listen to modern popular recordings at 78dB - since they're all compressed as to their dynamic range. This will explain the 6dB bass shelf from 105dB down added into the Harman headphone curve. Since I tend to listen at 83dB, I found it bassy.

But, easy solution from http://autoeq.app website, where I could remove the bass shelf and download a new target - which I am using with my various headphones ("...my hobby is collecting headphones"!)

Lastly, to the person who added it, love that link to Dr Sean Olives July 2023 post, thanks!

I am researching to confirm if my theory above is right...

Enjoy listening, everyone!

For like 30 or 40 years there have been audio amplifiers with a 'loudness' button toggle which applied corrective EQ specifically for listening quiet or loud.

The problem is with the fact that you cannot (you can guess) assume what loudness the person mastering used (=which EQ is correct). I don't quite understand the reasoning behind 'This will explain the 6dB bass shelf from 105dB down added into the Harman headphone curve.', which smells of elitism & assumption that the person in the Harman's experiment were 'plebs listening to modern records'. This is likely very untrue and hardly has any grounding. Hope you don't mind this digression; not trying to be hostile here!

FYI you would be hard-pressed to assume that people listen at 78dB - many people grew up on Apple iPod earbuds (or open on-ear headphones akin to Walkmans), which needed to scream over the noise, no ANC, passive reduction etc. In general, loudness war is one thing, but correctly applied compression makes or breaks the recordings.

The bass shelf screams to me "house curves" & the fact that many people even on ASR don't understand that speakers (non-mastering nearfields) are seldom setup to sound like a flat line, nor are recordings done with an assumption that the listener has a flat-lined speaker sound. Speakers don't usually sound flat, just take a look at what curve is used for amirm's speaker tests.

The answer would be... a simple bass-shelf regulator included in headphones, but we can't have nice things it seems.

 

[markanini]

For like 30 or 40 years there have been audio amplifiers with a 'loudness' button toggle which applied corrective EQ specifically for listening quiet or loud.

The problem is with the fact that you cannot (you can guess) assume what loudness the person mastering used (=which EQ is correct). I don't quite understand the reasoning behind 'This will explain the 6dB bass shelf from 105dB down added into the Harman headphone curve.', which smells of elitism & assumption that the person in the Harman's experiment were 'plebs listening to modern records'. This is likely very untrue and hardly has any grounding. Hope you don't mind this digression; not trying to be hostile here!

FYI you would be hard-pressed to assume that people listen at 78dB - many people grew up on Apple iPod earbuds (or open on-ear headphones akin to Walkmans), which needed to scream over the noise, no ANC, passive reduction etc. In general, loudness war is one thing, but correctly applied compression makes or breaks the recordings.

The bass shelf screams to me "house curves" & the fact that many people even on ASR don't understand that speakers (non-mastering nearfields) are seldom setup to sound like a flat line, nor are recordings done with an assumption that the listener has a flat-lined speaker sound. Speakers don't usually sound flat, just take a look at what curve is used for amirm's speaker tests.

The answer would be... a simple bass-shelf regulator included in headphones, but we can't have nice things it seems.

You're showing your ignorance. The effect of listening at different SPL changes the perceieved frequency responce. It's a well researched topic, not an invention of amp manufactures. Anyone can experiment with the values in csglinux's graph tool and see how it changes the target curve.

 

[GaryH]

They're not the one showing ignorance.

 

[ZolaIII]

Does it need to said? You don't have to question Harman as a baseline to have a valid reason for diverging from Harman for personal use.

Really? Hear you go a couple of baselines to think about (for speakers).

Reality is that exact universal baseline doesn't exist, there are more or less approximate one's depending on desired SPL and hearing state and that's all there is to it.
Hire is ISO 226 2003 to a baseline and it does make a difference depending which one you chose.

Equal loudness normalisation resources (a lot of them) all do not equal are consistent and together form much larger experimental base than anything Harman ever did.

Don't get me wrong I don't have anything (efficient) against Harman and it's totally OK with me if you like it for a baseline but it's entirely wrong to promote it as final solution or exact target. Such a thing doesn't exist and probably never will, at best you can get closer towards your personal taste (again related to hearing state and desired SPL towards used analog transceiver [driver] used) and that's about it.

 

[markanini]

final solution

That escalated quickly.

 

[Feelas]

You're showing your ignorance. The effect of listening at different SPL changes the perceieved frequency responce. It's a well researched topic, not an invention of amp manufactures.

That was exactly what I intended to show (that it's a well-known effect for years) through this figure of speech & inferring "amps had it for years => it is well known".

 

[Feelas]

Really? Hear you go a couple of baselines to think about (for speakers).
View attachment 303862
Reality is that exact universal baseline doesn't exist, there are more or less approximate one's depending on desired SPL and hearing state and that's all there is to it.
Hire is ISO 226 2003 to a baseline and it does make a difference depending which one you chose.
View attachment 303866
Equal loudness normalisation resources (a lot of them) all do not equal are consistent and together form much larger experimental base than anything Harman ever did.
View attachment 303867
Don't get me wrong I don't have anything (efficient) against Harman and it's totally OK with me if you like it for a baseline but it's entirely wrong to promote it as final solution or exact target. Such a thing doesn't exist and probably never will, at best you can get closer towards your personal taste (again related to hearing state and desired SPL towards used analog transceiver [driver] used) and that's about it.

Exact target no, but a suiting, universal approximate of actual "speakers + HRTF" shape is a usefil approximation. Much better at it that BD's "high treble kinks" or different nonsense curves like Grado etc.

The rest of FR shape (apart from smoothness + non- peakiness) barely matters in age of GEQ/PEQ/convolution unless realtime latencied are desired.

 

[markanini]

Much better at it that BD's "high treble kinks" or different nonsense curves like Grado etc.

Maybe Beyerdynamic back in the day made their concession in favor of efficiency, at the expense of stock tuning, because they envisioned the listener plugging into headphone outputs of a receiver, or mix console, which came with built in tone controls.

Maybe it's not as much Harman changed that paradigm as the increased accessibility of quality dedicated amps, commonly without tone controls.

 

[ZolaIII]

Exact target no, but a suiting, universal approximate of actual "speakers + HRTF" shape is a usefil approximation. Much better at it that BD's "high treble kinks" or different nonsense curves like Grado etc.

The rest of FR shape (apart from smoothness + non- peakiness) barely matters in age of GEQ/PEQ/convolution unless realtime latencied are desired.

Digital signal processing has its limits towards limits of actual analog driver, room, front to back reverberation... The age you are referring to with transparent (to our hearing) digital to analog conversion (24 bit PCM with SINAD above 100 dB) with potent DSP actually started some 20 year's ago. I played a lot with it past 20 year's including complex convolution kernels and such.
@markanini no not really (a escalation) just my own opinion with which of course you don't have to agree. Anyway point is it's not as simple as that (it never is).
Best regards and have a nice time.

 

[Feelas]

Digital signal processing has its limits towards limits of actual analog driver, room, front to back reverberation... The age you are referring to with transparent (to our hearing) digital to analog conversion (24 bit PCM with SINAD above 100 dB) with potent DSP actually started some 20 year's ago. I played a lot with it past 20 year's including complex convolution kernels and such.
@markanini no not really (a escalation) just my own opinion with which of course you don't have to agree. Anyway point is it's not as simple as that (it never is).
Best regards and have a nice time.

With regards to speakers - of course. As for headphones, the field looks like it's barely being explored, honestly which is precisely why the Harman discussion looks strange. As such, DSP limits in HP still looks hard to assess, discuss even. It's because the headphone research in general is neglected to an extent where anything more 'objective' stirs the brew.

I'd call that successful consumer conditioning.