First Test of Estimated Harman Target Response Curve on Various Headphones

As many of you know, Sean Olive and crew over at Harman International are developing a new target response curve for headphones. In this post I've modified InnerFidelity headphone measurements on a number of widely known headphones in order for you to have a first look at what the frequency response of headphones might look like with this new compensation curve.

The curve I used to compensate for the headphones is shown at the top of this page. The Harman curve (or a preliminary version of it) can be seen below.

140202_Blog_HarmanResearchUpdate_GraphDFvsOliveWelti

I did modify the curve slightly, however. Above about 5kHz, when I used the Harman curve for compensation, I found the response in that area to be significantly elevated relative to what I hear on the various headphones. I therefore changed the curve a bit by elevating this region about 1dB to 5dB as the response moved from 5kHz to 20kHz.

I then made copies of a number of headphone measurement spreadsheets for a number of popular headphones with which enthusiasts will likely be familiar. I'd very much appreciate your feedback about the compensated frequency response plots on this page and how well they match with your listening experience.

Without further ado, I present a crude rendering of what compensated headphone frequency response plots might look like in future.

140321_Blog_HarmanResponseFR_ATHM50

140321_Blog_HarmanResponseFR_BWP5

140321_Blog_HarmanResponseFR_D5000

140321_Blog_HarmanResponseFR_DT880250Ohm

140321_Blog_HarmanResponseFR_ER4PT

140321_Blog_HarmanResponseFR_FocalSpiritProfessional

140321_Blog_HarmanResponseFR_GradoSR60

140321_Blog_HarmanResponseFR_HD600

140321_Blog_HarmanResponseFR_HD800

140321_Blog_HarmanResponseFR_K267StudioSetting

140321_Blog_HarmanResponseFR_K712

140321_Blog_HarmanResponseFR_KossPortaPro

140321_Blog_HarmanResponseFR_LCD2

140321_Blog_HarmanResponseFR_LCD3

140321_Blog_HarmanResponseFR_NADVISOHP50

140321_Blog_HarmanResponseFR_ShureSE535

140321_Blog_HarmanResponseFR_SR009

140321_Blog_HarmanResponseFR_VModaM80

Please note that one of the reasons I did this is I'm in the midst of reviewing the Audeze LCD-X and found the sound of their headphones tipped up in the mid and upper treble more than I've experience in the past. I also had a hard time feeling they had enough bass after being exposed to the Focal Spirit Pro and NAD VISO HP50, though they're obviously well extended. Anyway, I got suckered into this rabbit hole and found the exercise illuminating and though you might feel likewise.

What do you think?

COMMENTS
funkmeister's picture

Wow! I'm digging the HP50 response. Also, I'm a fan of driver stabilization techniques like their shmancy ring on the outer diaphragm edge.

Wait a second... your measurement gear and Sean's are different. You need to pack up your gear and bring it to their reference room to get the curve nailed.

Remember, though, that it was a JBL room... a company known for great bass and pros at the MOAR bass fad... and some sharp treble. So, I'm not surprised at the compensation in the curve that they needed to make for headphones.

Keep this stuff up.

 

p.s. I tuned my EQ compenstion for my headphones to the curve as best as I could and I absolutely love it!

funkmeister's picture

I should probably point out that I had read everything I could on this leading up to this article, so I have the curve quite similar to the O-W line, not Tyll's line, but I actually followed what the test subjects had done with the bass and therefor my curve does not have a +5dB bass shelf but rather a +2.5dB shelf. That's how I interpreted the data and how my ears agreed. The shelving filter is the same kind of gradual (I used 0.65 Q) with its center point at 120Hz so the curve looks like a rise starting from ~180Hz and capping around ~70Hz. Otherwise it follows the curve.

I should also point out for the sake of readers that my EQ looks nothing like the curve because my EQ is setup to compensate my AKG headphones' ear-drum measurement to follow the curve. It took some time doing math to get it right.

ultrabike's picture

To me things look a little bass lite... A bit more so than what I think I heard from some of those cans. This might be due to the extra 5 dB in the bass region of the target curve. I think the bass compensation comes from perception tests, but it still seems somewhat arbitrary to me. Not sure there is a theory justification for that as there is one for the HRTF compensation for the mids and treble region.

Also, does Harman use a similar Head Accousitcs HMSIL3 measurements head/torso rig as IF?

I Like Headphone's picture

Tyll, with all due respect, I don't think you should be modifying the target curve to meet your personal preferences.

What was the point of all that research if you are just going to change it?

 

Looking at the sample graphs, it is very unclear to me where the target curve is among the measurements. Is it still supposed to be a flat line?

I think it would be clearer if you were to use a thick black line for the ideal target response, and somewhat transparent colored plots overlaying it for the measurements.

Jazz Casual's picture

x2

Dano91's picture

x3

starstarman's picture

x4

IMHO, the golden ears FR graph is easier to interpret and feels less subjective (although it was intended to be quite a subjective review) now. I'm not sure but when one says their target compensated graph could change, it feels like he means more preference than reference. I truly hope not.

Tyll Hertsens's picture

I was indded reluctant to change the curve, but when I used my closest reading of their curve above 5kHz the treble on everything was simply out of whack. But I understand your concern and agree that it's troublesome.

Basically, the raw frequency response plots (the gray set below the compensated plots) should look like the target response curve, and the compensated plot should be a flat line for a perfect headphone. SOmeday I might be able to produce plots with a target curve in the background---this time around is just a quick and dirty look at what headphones look like with a new target response curve.

purrin's picture

Tyll,

Everything looks too bass light now. I like what the curves did with the treble - they seem more reflective of how I hear things. I'd say keep your original bass compensation from 300Hz below but use the Dr. Olive's curve from 300Hz up.

This is based on my own experience building speakers (two in the past year - and a dozen sitting in my parent's garage in the past twenty), and now making a lot of use of digital parametric EQ to get the right response target.

In any event, you don't know how Dr. Olive's setup is like compared to yours. It could be probably (and I suspect) that the methods used in getting a proper seal on the headphones could vary. 

As we have discussed, getting the right amount of seal is a little bit of an art (not too much and not too little) with headphone measurements.

Cheers!

Marv

www.changstar.com 

Stefraki's picture

The headphone community may be used to a flat line between bass and mids as "neutral" but simple listening tests (such as online tools with test tones across the frequency range at different volumes) should be able to prove to you quite simply that they are not. Once you get close to 100hz, the tones get perceptually quiter.

In real life our tissue and bones play a huge role in the perception of low frequencies, which headphones cannot provide. Measuring of speakers and headphones does not compensate for the fact that our bodies and bones multiply the frequencies below 100hz quite considerably.

By that logic, headphone should have shelved up bass for it to sound even to us, IEMs in fact need it shelved up even higher as the outer ears, and small portion of the skull, involved in headphone listening are also removed from the equation.

People's minds adjust bass level to what they are used to, so I am sure a lot people will be up in arms about their favourite "neutral" headphones being revealed as bass light, but simple perceptual listening tests will show this is clearly the case.

I've always trusted perceptual listening tests more than I did graphs, because the two just don't match. With these new curves, they actually now look closer to what we really hear.

Tyll Hertsens's picture

"In real life our tissue and bones play a huge role in the perception of low frequencies, which headphones cannot provide. Measuring of speakers and headphones does not compensate for the fact that our bodies and bones multiply the frequencies below 100hz quite considerably."

If you read the Harman papers carefully you'll find they thought this might be the case, but in further experiments they found that people were selecting about 2dB LESS bass in headphones than on speakers. Totally counterintuitive, but there you have it. None the less, I do agree that a +2 or 3dB bass boost below 120Hz is likely proper for pleasing listening.

Stefraki's picture

They want less bass on speakers than headphones because when they listen to speakers the low bass is augmented by body absorption.  The reason they need a boost in low bass with headphones is because that absorption is missing.

Makes sense right? 

Alondite's picture

Read what Tyll said again. He said that the guys over at Harman expected that headphones would require a bass boost to sound neutral, but found that people actually found that headphones sounded neutral with 2 dB LESS bass than speakers.

What makes sense logically doesn't always work out that way in practice. 

Stefraki's picture

The graphic Tyll uses to show the speaker and headphone curve has the speaker's bass as lower,  which is where I was coming from.

Regardless,  I did notice what Tyll had said after posting and said so a minute later in the follow up post.

Stefraki's picture

You're right that is counter intuitive, I wonder why it is...

Tyll Hertsens's picture

I guess personally I think the +5dB bass is a bot too much, but I do think maybe +2 or 3dB might be good. I think flat is a little bass light. 

The new curve looks a LOT better to me from 1kHz to 8kHz where with my normal compensation curve regularly gives a dip in the response. So I like that a lot.

I would assume the head they use is IEC spec and sould be quite close in response to mine. But yeah, getting a seal for bass response is fairly tricky and headphone mounting proceedures definately might be a source of differences in bass response between their measurements and mine. 

Alondite's picture

I'd definitely have to agree with you here, Tyll. I think +5 dB makes headphones a bit bass-lite, but  +2-3 dB is probably appropriate. 1-8 KHz does seem a lot better than the previous curve, but it seems slightly elevated on the higher end. Above 8 KHz seems to be even moreso. 

audioops's picture

"I guess personally I think the +5dB bass is a bot too much, but I do think maybe +2 or 3dB might be good."

"The new curve looks a LOT better to me"

"So I like that a lot."

You seem to be swinging between being totally arbitrary and skewing your results towards where you most wish they were. I've read some of your thoughts about the limitations of objectivism in headphone analysis and so I assume I have a fair idea of your feelings about the limitations of measurements in general, but this tampered analysis is truly unhelpful. I would love to read your thoughts about how good/bad different headphones sound to you compared with how research predicts they will sound.

BnKbrainstorm's picture

+10

Amen

TMRaven's picture

The +5 bass boost on Harman curve is a tricky one.  On some headphones I feel the 5db of bass boost was mandatory (focal spirit pro), but on other headphones like the LCD2 and HE-400, I feel as though they extend deep down well enough without needing any bass boost whatsoever.  Part of this might be the natural abilities that Planar Magnetics have in bass extension, part of it might not.  Then there are some closed headphones like the M50, which are definitely way above neutral as far as bass goes. The fr graphs can only show so much, and while a closed headphone with high thd and long decay might look like it has a bass rolloff on the fr graph, it might be subjectively bass heavy.  

I agree with Purrin, I say use the harman curve for the midrange and treble, while keeping your original compensation for the bass frequencies.  The new midrange-treble compensation here goes better with my subjective experiences with some of these headphones as well.  With this new curve applied to the midrange and treble I feel as though we can actually consider flat on your grpahs to be pretty close to flat in response.

Tyll Hertsens's picture

" With this new curve applied to the midrange and treble I feel as though we can actually consider flat on your grpahs to be pretty close to flat in response."

Indeed. Seeing that area come out basically flat (from what I can tell) is very satisfying. Very greatful for Harman's work on this.

Claritas's picture

I hope you don't add anyone's curve to your charts as people don't need to be told what to like and it also clutters the view (already busier than some other charts).

Jazz Casual's picture

It's not about telling people what to like. It's about establishing a new reference based on scientific research that is more relevant to the headphone listening experience than the current one. Placing the new headphone target frequency response curve on the measurement chart (as Golden Ears does with its frequency response measurements), would make it easier for the reader to compare how the measured frequency response of a headphone compares to the ideal target curve. This could be achieved without making the frequency response chart look too busy.

Tyll Hertsens's picture

When we get a target resonse curve we believe in, an ideal headphone compensated would be a flat line. That's the whole point of the curve: to be able to assume ideal is a flat line and deviations from flat is a coloration. How is it hard to interpret that? 

If you're talking about an ideal curve behind the uncompensated plots, that might be useful, but if I'm showing the compensated plot as well why would you need it?

Jazz Casual's picture

If a flat line is going to represent the revised target response curve then no, you wouldn't need it. I was referring to what you've presented here for consideration and agree with I Like Headphone's comment regarding the appearance. The Golden Ears charts came to mind, which provide the reader with its target frequency response curve and the actual frequency response measurements for comparison, which I think is a helpful feature. 

Claritas's picture

How does this "scientific research" ultimately differ from taking a poll, however sophisticated its method? How does majority preference demonstrate that something is ideal (or true, good, beautiful, &c.)? Why is the opinion of several men better or worse than that of one? I hope to hear some thoughtful responses because so far an ideal curve sounds like a compete boondoggle.

Tyll Hertsens's picture

Harman is looking for a sound that pleases humans. The only way you can do that is to use humans as subjects. The difference between testing one and many is that we're a messy lot, and our opinions vary widely, so you have to test a lot of us to get an idea of where the center is.

The actual problem is that there is no adopted industry standard for studio speakers either, so we never know what the recording engineer was listening to when he or she mixed the music. So your thought that this whole thing is a boondogle actually misses how big a problem we have.  Floyd Toole calls this something like "the circle of confusion." Sit back and relax, we're going to be watching this problem for a long time.

AstralStorm's picture

The main problem with Harman's study is they didn't use the recommended MUSHRA quality trial practices, only left people tweaking and recorded the result. This is broken due to various psychoacoustic reasons. (Such as getting used to a particular sound.) Plus the sample size is way too low for any conclusion.

A true quality trial would have a bunch of settings and people giving 0-5 quality scores to them, with a few so-called "anchors" - for instance: classic diffuse field target, free field target, no compensation, various modified diffuse and free fields, AM radio, cellphone.

While it is true that most IEMs and some headphones are missing a bit on subbass, their curve is starting the boost way too high - their shelf equalizer was just set this way. The upper mids are indeed bit too high to my ears - similar to original Etymotic/diffuse-field curve.

My theory about the sound is the people were comparing to speakers in a room - most rooms have standing waves in the low end - more than those tiny 2 dB they set less on speakers.

If we were to hold to high fidelity target, the real benchmark would be an accurately recorded (as in measurement microphones) live performance. This means as close to vanilla performance as possible, making any mastering issues moot.

BnKbrainstorm's picture

Nothing has been proven to be more subjective than sound. 

OK, that's a bit of a stretch but you get my point.

Ideal curves are important and useful design tools but do very little to account for the endless permutations of filter arrangements present in the vast sea of human heads.  Anyone that has immersed themselves in the acoustics and pyscho acoustic fields of study know very well that this mumbo-gumbo about "ideal" curves from a listening standpoint is fodder for endless debate.  And if it's debate you seek to spend precious lifecycles on, then bravo, you've found your utopia. No two people perceive pure tones much less complex, dynamic frequency spectra alike. So in other words, the "ideal response curve" can not be a replacement for your own response curve.  And your own response curve is anything but static or fixed; it shifts. and sometimes dramatically.  This is why there are those things called Tone Controls & EQ's.

Loudness, peaking, distortion, phaseshift, tonal stack, etc all come into play and like finger prints, no two head superstructures are alike.  More importantly, add to it the complex human brain and all bets are off with regard to "ideal" response curves and the perception of such therein.  Recognize these variables and the pursuit of the the "ideal" becomes moot.  Perhaps "ideal" guideline makes more practical sense.  Perhaps not.

The perception of sound spectra fitted to some "curve  du jour" (A-weight or otherwise...) doesn't matter much past the excitement of the debate.  What matters is the trust you have in your own ears.  Couple that trust with your best sensibilities on choosing function & comfort and viola! you have your ideal headphones.  Interestingly enough, Tyll decided to massage the curves based upon his own perceptions.  That alone should show you that this perception dance is "subject" to interpretation and arbitrary.

"Ideal curves" are like sea buoys; (in this instance) they give audiophiles something to home in on, a marker if you will and something perceivably tangible to debate.  Don't get me wrong, points of reference are highly valuable, but just like predictive modeling in any engineering discipline, they're imperfect and changing based on a constant stream of newly discovered input variables.  And in this instance, "subject" to a vast array of personal preference and opinion.  So juggle me that Batman...

Jazz Casual's picture

"Nothing has been proven to be more subjective than sound. OK, that's a bit of a stretch but you get my point." Is it ever. Is sight, touch, taste and smell any less subjective? The Snellen chart is used to test visual acuity for the general population. Optometrists test the eyes of countless individuals using standardised testing. Blood pressure, respiratory, kidney and liver function, cholesterol and blood glucose levels are all tested against standards developed for use on the general population. I'm sure that the more medically minded amongst us could cite numerous other examples. Most relevant though is the evaluation of our hearing by audiologists using standardised testing procedures. So riddle me this, what makes the development of a standardised target curve for measuring headphone frequency response any more challenging? ;)

BnKbrainstorm's picture

Because hearing tests primarily intend to do one thing with respect to broadband spectra, octave band and more importantly, tone detection.  The simple answer:  To verify audibility or the absence of sound.  There is significant margin area with respect to binary tests.  Completely irrelevant to the debate at hand. 

THIS debate is about "sound quality".  And sound quality is both highly subjective and variable depending on the angle of incidence, the shape of the ear & its associated components and the brain processing the sound in question.

Standardized response curves are dandy for mfging control and generalized marketing.  But the massaging of curves, a dB here, +3 dB there, is laughable on broad band spectral content (with respect to qualitative analysis...).

Jazz Casual's picture

 What is the basis of your claim that "No two people perceive pure tones much less complex, dynamic frequency spectra alike."? Are these differences in aural perception to which you refer minor, significant or somewhere in between? If the magnitude of these differences were great for the population at large then musicians for example, should find communicating with each other far more challenging than they do. Also, the term "sound quality" is a loaded one and I'm not sure if that really is what this discussion is about. Dr Olive's team has proposed a target curve for measuring headphone frequency response based on research conducted on test subjects. A verifiable trend has emerged despite the variables associated with people's hearing and their personal preferences that is the foundation of your subjectivist position. 

BnKbrainstorm's picture

Postulate flawed:  "If the magnitude of these differences were great for the population at large then musicians for example, should find communicating with each other far more challenging than they do."  Define "great" in engineering units.  Better yet, define the critical, discreet engineering units that provide the "sounds good" factor much less provide granularity to that mystical link or connection.  It's been tried.  Not possible.

The basis is simply common sense and the indepth study of acoustics in the industrial/consumer sector.  By discipline, I'm a mechanical engineer who's been involved with and responsible for countless subjective listening tests in response to the OEM integration requirements for various electro-mechanical devices in both the consumer space and the industrical sector.  I would suggest you do some indepth reading in the subject matters of third octave weighting schemes + human anatomy; specifically the inner ear cochlea.  Then perhaps you'll understand the whole, no two alike, finger print analogy.

Let us know when you have a scientific method for exacting just how sound is perceived between human subjects, not just to the dB level but to the actual shape of perception, good - bad - or otherwise.   And yes, although agreeably "loaded", you simply can not deny that sound quality is the thrust of this particular exercise in futility and the premise (goal) of measuring/generating the response curves in the first place  -  in this particular thread stream and in the promotion of headphones generally.  Response curves provide excellent engineering guidance (A vs B), an equally valuable point-of-visual-reference for end use comparison but at the end of the day, you still rely heavily on human feedback, and trust me, it's all over the map for obvious reasons.  Reasoning that stretches way beyond the so-called "connection" musicians make when being creative.  Try applying scientific method measurement philosophies to that relationship Jazz...

Jazz Casual's picture

I see this kind of response a lot in online forums. I can deal with the condescending attitude, the attempts to blind with science and the need to resort to stating one's expertise to make assertions appear more credible. I am not a mechanical engineer but I know techno-babble and disjointed assertions masquerading as counter-argument when I see it. With regard to your "no two alike, fingerprint analogy", yes of course we are all different but we also have commonalities. It doesn't necessarily follow that we all should have significantly different sensory perception as a consequence of being different individuals. I don't accept that the thrust of this thread and our exchange is actually about "sound quality". If you read Tyll's earlier article on the research being undertaken by Dr Olive and his team, you will see that they are attempting to establish a target curve that listeners will find most pleasing, which is arguably quite a different thing. And I do not agree that it is an exercise in futility and doubt that Dr Olive's research team would either; nor have you proven it to be so. This disagreement has become circular as they tend to do and frankly, rather tiresome. 

BnKbrainstorm's picture

I'd recommend that you don't pose questions regarding "basis" then, Jazzer.  "Pleasing"?  Absolutely nothing to do with sound quality you say.  "I don't accept..."  Unbelievable.  But too funny.

This is almost better than a Global Warming debate. 

Audiophiles; what a hoot! gotta love em!

Jazz Casual's picture

That response doesn't cut it either. My questions were perfectly reasonable, using acceptable language that we all understand, and you failed to address them for the most part for the aforementioned reasons. And please refrain from putting words in my mouth. I didn't say that a pleasing sound has absolutely nothing to do with sound quality. I said they are arguably different things. For example, a pair of highly resolving headphones or speakers, may not sound as pleasing to the ear as a less resolving pair. However, improvements in sound quality should enhance the listening experience and hopefully make it more enjoyable. I refer you again to Tyll's previous article on the research being done by Dr Olive and his team. I've noticed that experts in their fields are often humble, and have a great ability to explain highly technical concepts and complexities in layman's terms - qualities that are sorely lacking in your posts 'brainstorm.

arnaud's picture

I agree that the huge variations in sound pressure at the ear drum resulting from large variations in morphology across the population do not necessarily relate to perceived differences sound quality or even loudness.

By that I mean that our brain is intimately tuned to our morphology, to the point he can even compensate for hearing loss to some extent with the person not even realising he/she's getting deaf in the midrange for example.

It's actually very interesting topic as it questions the need for very peculiar / detailed coupler for headphone measurements... Basically, and I think I read dr. Olive mentionning this himself in discussions somewhere on another forum, the coupler main goal should be to provide a proper acoustic load to the driver. The peculiarities of the ear shape and such should partly cancel out during the application of compensation curve, hence you could argue it doesn't matter how peculiar an ear shape you start with. It just have to have approximate dimensions of a population average, the existing standard seems to comply just fine.

Arnaud

 

Hjelmevold's picture

I recently turned 30, so I'm still young enough to have good hearing up to 16-17kHz. And I can safely say that the compensation outlined here must be too dramatic above 10kHz compared to my experience, portraying headphones that have very poor response above 10kHz as if they are nearly flat or even treble-heavy.

I have the LCD-2 myself, and I'm certain that there is no 15-20dB gain from 10-20Khz! Here's how I can tell:

Suppose I sit in the sweet spot of my studio monitors (PSI A215-M), with my LCD-2 headphones on, powered through my Phonitor with crossfeed at settings that simulate my monitors. The studio monitors have been DSP-calibrated to a flat frequency response, verified in RoomEQWizard. Then, I set the volume of both my monitors and headphones to equal loudness, which I can verify by hearing approximately a 3dB drop if I mute one of them on my monitor controller. Then I run a slow sine sweep. What do I hear? The sound of the LCD-2s disappears once the sweep reaches about 2-4kHz, never to be heard again for any higher frequencies. Meaning that these headphones have sufficient amplitude loss in these frequencies to be masked by the sound from the studio monitors, requiring at least a 6dB drop.

I can confirm this in another way as well. I have a set of in-ear binaural microphones: SoundMan OKM II Studio. The nice thing about these, is that since they mount in the concha of my outer ear, they can be used as a reference to measure both speaker and headphone response. Sure, the microphones are diffuse-field equalized to be compatible with playback on speakers, so as a result they don't have a flat frequency response. But what is more interesting here, is the difference in a recorded sound sweep done on speakers and on headphones. When comparing these measurements, I actually end up with a response curve pretty close to the current HRTF compensation used in today's measurements on InnerFidelity. So I don't agree with these new response curves, at least not if they're supposed to qualify as an HRTF curve.

Another point worth mentioning, is that I can test the response curve presented in this article by EQing my flat studio monitors, so they approximate the difference between the dotted green line and the black line in Figure 2 of this article. In other words, a gradual 6dB bass from 200Hz and below, and a gradual 3dB drop from 800Hz-4kHz, then flat -3dB above that. And as a result, my regular clients who come into my studio and listen to my monitors with this EQ, comment that my studio monitors "sound so modest and natural". But I know that they don't sound flat anymore.

So it seems apparent to me that this target response curve is designed to match the color of when recorded music sounds natural. Not to be a new type of HRTF compensation that correlates to a flat frequency response. The former may be preferable if you want headphones for music playback, while the latter may be preferable if you need to use headphones as sound referencing tools.

Tyll Hertsens's picture

...the LCD-2 headphones used in the above measurements was one of the brand new ones that has significantly elevated upper-treble relative to the older ones. So your experience of your LCD-2 is unlike the ones shown above.

arnaud's picture

I'd be curious to see the 009 response without the 5dB boost being enforced by the new compensation curve. I think it'd look pretty darn flat... 

Now, since some have some issue with a "bass light" nature of the 009 (but some don't, I am one of them), I think it's pretty clear there won't ever be a curve to fit the bass heads, treble heads and other peculiar ears out there...

Also, I can't make sense of this considering the target compensation curves are basically tied to the measurement head. In particular, the compensation above 2kHz is highly influenced by the ear shape and acoustic impedance of the surface material used. Tyll, your current comp. curve is for your type of head and some number of incidence angles (never been quite able to figure what headings were there for this "Independent of Direction" compensation). How does the Harman curve become applicable to your measurements, are they using the same head?

If so, or even if not, it does bring the point about the variability among the population. In particular, I am not sure we can ever define a single curve because of the morphology issue pointed above or even simple fact we don't all define neutral the same way. On the other hand, a good statistical representation of the target population and some brakets to fit within that population to a certain percentale would make a whole lot more sense to me.

Arnaud

Tyll Hertsens's picture

"Also, I can't make sense of this considering the target compensation curves are basically tied to the measurement head. In particular, the compensation above 2kHz is highly influenced by the ear shape and acoustic impedance of the surface material used. Tyll, your current comp. curve is for your type of head and some number of incidence angles (never been quite able to figure what headings were there for this "Independent of Direction" compensation). How does the Harman curve become applicable to your measurements, are they using the same head?"

That's a potentially a big problem. On the one hand, if they're using an IEC spec head the ear shape should be identical to mine and it's response quite close---that's the reason for industry standards after all. On the other hand, it wouldn't surprise me a bit that their head is slighly different than mine so differences certainly could exist.

Their target curve is quite smooth and I've got to think it's not exaclty what they measured on their head but rather a simplified look at the basic curve shape. It's absolutly true that if I put my head in a well designed room with ideal speakers, the response plot would have much more detail than the compensation curve I've shown at the top of the page. So there's no doubt that some of the wiggles in the plots above are due to the artificial head's response and not coming from the headphones. There's no doubt in my mind that aquiring a compensation curve by putting my head in a proper room and recording the response on that head would be the ideal way to go....but it's a rather difficult thing to pull off. Sure would be fun though.

"If so, or even if not, it does bring the point about the variability among the population. In particular, I am not sure we can ever define a single curve because of the morphology issue pointed above or even simple fact we don't all define neutral the same way. On the other hand, a good statistical representation of the target population and some brakets to fit within that population to a certain percentale would make a whole lot more sense to me."

My understanding is that the IEC ear is very carefully developed by looking at a wide range of human ears, so your concern is at least somewhat satisfied.

 

*sigh* We've got a long way to go yet.

arnaud's picture

Tyll, as per the <<*sigh* We've got a long way to go yet.>>, it seems rather to me like you've come a long way. From the overall feedback here, it seems like this new curve you are using based on Olive's work would be working well, maybe some adjustment at LF only.

I take note of your point about the IEC ear fitting a population average (the picture you've shown of your head with very human like ear shape did not seem like like one of those average geometry but very lifelike instead).

As for fitting all tastes, I found I could very much trust your hearing so I'd say, don't change anything, you're done some of the best job out there as independent gear reviewer and measurement lab!

Finally, you probably don't recall the experiment but I had looked at de-convoluting the inverse of your compensation curve (assuming minimum phase characteristics because apparently headphones responses are very much so for the most part) from the impulse response and other time domain quantifiers you provided on the measurement sheet. The result was certainly interesting with the 009 (and other well measuring phones like the HD800) getting much improved step and impulse responses. This would make sense as the compensated frequency response is reasonably flat in both instances, and a flat frequency domain response with reasonable bandwidth would most ressemble an "ideal" stair case for step response in the time domain. I recall impulse response was cleaned a bit also.

So, assuming you can move forward with the plan to use a revised target curve, having it applied to the rest of the tests would seem like a great move forward. Of course, it won't ever be perfect (for instance the minimum phase assumption) but, imo, it would make for easier interpretation of time domain results. For instance, I am a loss to interpret the current step responses. You've seen / memorized many so you can do it well apparently but maybe not so many of your readers.

Finally, I don't often comment here but wanted to thank you for the continously fun / interesting / thought provoking / enlightning content on your blog!

Arnaud 

Stefraki's picture

To me, they look much more realistic to me. I have never got why every headphone that Are clearly even through the upper mids/treble always had a huge dip there. It made no sense, and these graphs now actually reflect what we hear better. 

I agree with the bass shelf as well, as I noted in a post above, simple listening tests can show that headphones which measure flat between bass and mids clearly get quieter below 100hz.

Only thing I'd say is that more research probably needs to be done on IEMs. With no outer ear or skull absorption of lower frequencies,  I would argue that they need much higher shelved up bass to sound neutral.

I actually think Sennheiser's IE800 got it right. The bass doesn't sound boosted at all, it sounds right.

Tyll Hertsens's picture

WHen last I asked Sean what their future direction of research, he answered that developing an IEM curve is high on the list. You're spot on.

Jim Tavegia's picture

Surprising for the Grado and Sennheiser cans.  But, many others as well. I wonder how my AKG 701 would measure vs the 712?  

Ulrich's picture

As Principal Scientist and headphone designer at Harman, I would like to point out what Tyll is showing, is NOT the headphone target response curve that we are using to design new headphones at Harman. As part of our Lifestyle R&D group, we agree with Sean on the published target response (bass boost and treble tilt), but found over time that there are better, more direct methods to determine the perceived frequency response of a headphone, than using ear simulators. I can't disclose details because patents are being filed right now, but promise there will be good things coming to market in the near future.

Dr. Ulrich Horbach, Harman Lifestyle R&D, Northridge, CA

Stefraki's picture

All of this research can only be good for the headphone hobby.

Thanks for your continued research Ulrich.

Tyll Hertsens's picture

Do you have any light to shed on the steepness of the curve drop off above 5kHz? It seems very steep to me.

Jazz Casual's picture

Dr Horbach - so you are saying that Tyll's target response deviates from Dr Olive's. That is understood, but your position regarding Dr Olive's research is somewhat ambiguous. Are you saying that you agree in-principle with Dr Olive's published target response but not with the methodology used to measure the frequency response of headphones? Could you clarify please? Thanks. 

Ulrich's picture

Jazz Casual - As I said, we are using new measurement methods in headphone design that are not published yet. Whether I personally agree with Sean's research or not doesn't really matter here. But I think the discussion in this forum is good and helpful.

Ulrich

Jazz Casual's picture

Thank you for your reply Ulrich. What I'm trying to ascertain is whether your new measurement methods yield significantly different results to those used by Dr Olive's team or are you measuring different things? Will the new target curve be adopted by Harman? Also, is there currently an agreed or standardised methodology for measuring the frequency response of headphones?

AGB's picture

Dear Sir,

I think, judging by the curves and the two headsets (one IEM) I have here, Shure 535, LCD-2 and the new X not detailed above, the curves may mean something or maybe not that much. The Shure seems to be right on, needing a gentle bass boost; the LCD-2 a treble boost about 1.5 and 7K, albeit I do have the older model. All of these need some EQ, the 535 to raise the bass, the LCD's (both of them) I need to cut the bass in the two lowest octaves to flatten the middle and add some boost in the sparkle area. This EQ creates more resolution in the low bass for the LCDs, removing the lumpiness in this region.

I would imagine that just about everyone will have different experiences, as there probably is no such thing as a standard ear, or a standard EQ, no matter how we'd like, with good intentions, to standardize both. The geometry and impedance of each ear and the resonances within the canal may be different enough for each ear to cause a variation in individual response.

This problem is intractable I think and will cause disagreements, or better, an acceptance of reality - that all earphones need some EQ unless one is lucky and gets a device with which his own ear defficiencies counter than of the device.

Harman's research is great, may produce some good products, but I would not rely solely on it. I would rather consider the path taken by a few other makes wherein FR adjustments will be possible for the user or make EQ adjustments with DSP included with the better music player software.

I am however somewhat pleased to see that others here have caught on to the necessity of EQing that I have been suggesting for a few years, especiallly for those for whom no-EQ uber alles is an unbroken rule. Thinking that equalizing is a poison, they don't realize that it is actually the antoidote for. Lastly, they really don't know what they're missing by denying themselves their earphone's musical potential.

I've been extrememly successful btw, with the LCD-Xes. A tour de force, especially with a must sligh EQ with which one can make major gains in imgaing, transient speed, resolution and transparency.

AudioGuard's picture

Dr. Horbach, What could be more direct than measuring on a coupler?  Much research has gone into couplers, which properly load the headphones and are standardized so different people can measure and expect to get the same result.  Are you trying to make the measurement method proprietary?

MatGyver's picture

The new target response curves looks about right if you ask me. I have allways EQed 20-150hz by a few +dB`s on most of my headphones. Mainly Audeze LCD`s and Sennheiser HD 5,6 and 800`s. And often the treble 2khz-5khz too by a few dB too depending on the headphone. Typically Audeze, Aiwa and Nad rp18. Some were perfectly balanced like HD555 and HP50. And some too bright and required -dB EQ like the HD800.

Audeze and Sennheisers open models have a midrange (150hz-1500hz) elevation i really dont like. Audeze more so because of the recessed treble. Both sound unnatural to me.

I really think Harman is on to something with their newest target response curves, even though the Harman JBL Synchros "based on this curve" was an epic fail that never should have made it in to production. The LiveStage DSP at the touch of a button was a stroke of genius though and I would like to see more models with this feature.

Alondite's picture

The new curves look better than the old ones, at least, but something still seems a bit off. Relative to what I'm hearing, I feel like bass is under-represented in the new curves, and treble is slightly over-represented, particularly above 10 KHz where they seem quite elevated. 

RPGWiZaRD's picture

Agreed, below ~100Hz slightly too recessed and above ~10kHz slightly too pronounced. That would also be in-line with goldenears & headphone.com measurements for example which to me have the most closest to neutral flat line accuracy representation. I've EQ'd all my headphones in the past (I like some bass boost, around 7~9dB or so versus the average amplitude of the mids & highs on goldenears/headphone.com's algorithm) but as even mids and highs as possible (^-- shape) and the M-Audio Q40 measures according to goldeanears very closely to ^-- shape with unusually well balanced mids vs highs and is the first headphone I didn't feel I need to EQ at all, both my ears and Goldenears measurements seems to perfectly agree and some other headphones how I EQ'd those agree also pretty well with their graphs.

So I'd say below 100Hz it's like 3-4dB too recessed and about as much emphasized above 10kHz from the looks of it but from around 100Hz or so to about 10kHz looks fine. 

In my dreamworld I wish all sites used same compensation algorithms, whether the compensation used is accurate or not isn't as important as the ability to compare different sites' measurements with each other easier and they start complementing each other (providing the reader have a bit common sense and understand to align similarly and to know that there is still many variables in play like isolation on dummyhead, different headphone driver samples and equipment used etc all will give slightly various results). Until now I've always had to have in mind a guesstimate angled "downwards slope" above 1kHz when comparing a measurement of innerfidelity vs other sites.

Stefraki's picture

Is not in fact correct? Very little is going on there in music especially acoustic music. We're talking about the harmonics of cymbals, breath and air, that sort of sound.

i have often found these elements to be over pronounced, particularly in high end headphones where it is sold as "detail retrieval". When I look at the headphones like SR009 and HD800 with a shelf in the 10k+ region, it makes perfect sense to me based on my listening of those models.

I do wonder if all this calling for the curve to be changed under 100hz or over 10k stems from a desire for headphones which have widely been considered as neutral (but clearly never were in listening tests) to be "protected" from being revealed as coloured. 

Alondite's picture

Headphones with too much energy in the 10 KHz+ range come off as excessively airy and even sibilant, especially vocals, which become far too breathy. Pink noise listening tests also shine a spotlight on peaks in this area. The HE-400 is known to be elevated in this area, and it's plainly audible. 

It's not about a desire to "protect" anything, it's a desire for it to be correct. 

MatGyver's picture

Tried to EQ my speaker frequency response to the Harman target reference curve Audeze LCD2 curve, and it does give me alot of Audeze LCD2 rev.1`s vibes. Its almost like listening to my old pair of Audeze except im not. Facinating.

Frequency response from speakers that remind me of Audeze LCD2: http://s28.postimg.org/c9n2fydd9/audeze.jpg

lmader's picture

I carefully overlaid the Harmon curve on Tyll's measurements of the Hifiman HE-400. I was careful to match the db and frequency response axes. The image is here:

http://cogitoabsurdum.files.wordpress.com/2014/03/he-400-vs-harmon-ideal...

I then applied some conservative adjustments to the EQ based on that graph, and I think the result was great.  I basically boosted the bass below 100hz by a couple db, boosted the 2k - 5k range by a couple db, and pulled down the 11k by a couple db.  To my ears, this is an improvement on the HE-400s. 

It seems to me that if we have a credible ideal curve like this, that we can EQ our headphones to this and magically improve our cans. Is this a reasonable idea?

NickS's picture

I'm interested in measuring and EQing my headphones to this curve. Would an in-ear omni mic like this be appropriate for doing a frequency response measurement of an on-ear or over-ear headphone?

http://www.amazon.com/LOW-NOISE-IN-EAR-BINAURAL-MICS/dp/B002N6PAMM

Rillion's picture

Hi Nick,

I have a pair of those in-ear mics.  They do have very good sensitivity.  Using them for headphone measurements is quite a bit more trouble than you might guess.  The average of the left and right have a pretty flat frequency response, but there is up to 3dB imbalance between them that varies with frequency.  I ended up making a calibration curve for mine with a Dayton Audio EMM-6 measurement microphone.  

I found that even using them to compare one headphone with another, the agreement was not good in the treble with what I heard by ear--even disagreeing about which headphone had a stronger response at a given frequency.  The model of in-ear mics shown in the picture sit just outside of and to the side of the ear canal (they leave the canal open so you can hear while using them).  The right one in my ears does not sit as deeply as the left one because of slight asymmetry in the structure of my ears.  

I have found that to get results consistent with what I hear, I have to use in-ear mics that sit exactly in the entrance of the ear canal (which, again, the ones you are considering do not).  I rigged up some ear-canal blocking mics by ordering some 4mm diameter microphones from Digi-Key, soldering some thin headphone wires to them, potting the wires with an electrical insulating material, and gluing the mics to the end of some foam ear-plugs that I trimmed down to fit flush with my ear-canal entrance.  I used a small alligator clip as a tool to help insert them without handing the sensive element of the mic. Yes, I know I am nuts.

In any case, the mics you are considering (if calibrated) will do a good job of measuring headphones up to about 800 Hz, a fair job (+-2 dB) up to about 1.25 kHz, and above that I would not trust them at all.  They have some pretty impressive bass extension--much better than the mics I bought from Digi-Key which are only spec'ed down to 100 Hz.  Even the ear-canal blocking mics give some questionable results at 10 kHz and higher (6dB and greater channel imbalances with some headphones which I do not really believe---but it is pretty hard to judge loudness by ear up there anyway).

I also used the ear-canal blocking mics to measure my HRTF at 0 degress and +-30 degrees.  That too is also quite a bit more trouble than you might guess: you need to find a way to remove the effect of room reflections.  There is another long story to that ...

If you do decide to obtain some in-ear mics, you will need a recorder that can supply them with power.  For this purpose, I bought a refurbished "Tascam DR-05 Linear PCM Recorder" from Sound Professionals.  I am very impressed with this recorder.

arnaud's picture

I went the same route trying to get a binaural measurement rig on the cheap using my own head so that I could also possibly explore the wonderful world of personalized virtualisation as introduced "to the masses" by smyth research.

They actually do just what you describe: using micro-capsules inserted at the entrance of ear canal using foam tips. I reckon they are calibrated in factory though.

As for myself, I did try the Sound Professional in-ear mics (low noise newer version) to find out that the mismatch between canals exceeded 5dB over the whole audio range and the SNR below 100Hz was plain terrible. I think you could still do something with it for virtualization (where the head & torso effects really only play from 1kHz up) but as a headphone measurement rig, I would pass. Not mentioning the fact I never got reimbursed even after returning the product / sending numerous emails, way to go customer service.

Lately, I am giving it another go with the Roland CS-10EM which, despite the lower price seem to be much quieter than the SP ones and probably give more consistent as per how they mount. Never got around to process the measurements I took though, so jury is still out... 

I second the tascam dr-05, it's working great with these mics, nice cheap investment for those who want to explore binaural sound capture.

Arnaud

Rillion's picture

Thanks for the info.  I had heard of the Smyth Realiser but did not know that it included in-ear microphones for customization.  Pretty neat stuff!  Too bad the kit costs $3k.  Considering the time I've sunk into these projects, though, it might actually be worth that much to have a more streamlined process--I might demo it one of these days.  The homemade/hobby approach does however provide some satisfaction in having a more complete understanding of the whole process.  

After all this obsessing over frequency response, the ironic thing is that some headphones that have a rather uneven frequency response (e.g. Phiaton PS320) still sound pretty darn good to me. 

Mike, AKA Rillion 

arnaud's picture

I also used to think 3kUSD was too much for me but realising over a year has passed and I haven't got the time to really go anywhere with this, it's probably not worth trying to reinvent the wheel... The product is unanimously praised, you should not be deceived...

Stefraki's picture

If you have the opportunity, compare any IEM that measures roughly flat between bass and mids (PFE101, Shure SE535, Hifiman HE600 etc, there are others I haven't heard) with headphones that measure similarly (such as the LCD2 etc) -

- does anyone agree that they sound to have similar bass levels? 

For me, the bass on IEMs which measure "flat" sound much thinner and less prominent that the bass on headphones which measure "flat". 

This is really on the forefront of my mind as I recently got some IE800 and was worried I wouldn't like them due to the famed "huge low bass boost" but running them through my standard listening tests I found no such thing. Deep, even bass, powerful for sure, but equal in volume with the mids. Quite perfectly even in fact. 

I the reception is a combination between IEM people being used to hearing shelved down bass as "neutral" and the fact that people have just seen frequency plots with shelved up bass and decided that they have to go along with "the science". You often hear very wierd things written about them such as "the low bass is clearly boosted, but sounds balanced with the rest of the frequency response" - which is obviously a contradiction in terms. 

I would not be surprised at all if when Dr Olive comes out with his IEM curve Sennheiser are vindicated and the bass shelf on the IE800 is actually spot on. 

Alondite's picture

The IE800 are not neutral. Simple Pink noise listening tests clearly demonstrate that the low end is above neutral in quantity. There's even some bass bleed into the midrange, which is obviously a coloration. 

Comparing bass quantities on two different headphones is an exercise in futility because the quantity of the other frequencies relative to the bass are going to affect your perception of the bass. Low frequency harmonic distortion is also going to influence perception.

People have not been hearing neutral headphones as "shelved-down," bass-heads have been asserting that there is "not enough bass" in neutral headphones. I've heard people say that my HE-400s even have "no bass," which is ridiculous. 

How anyone could be a bass-head I'll never understand, because low frequency notes contain the least amount of musical information and have the least influence on the sound.

Stefraki's picture

Mine are neutral through the bass to mids. I've confirmed that with test tones. I'm not going to say they are neutral all over (treble has dips and peaks) but the low bass right into the mids remains even - with my set and my ear canals - your IE800 and ear canals may vary. ;-)

Alondite's picture

Test tones are not effective for comparing relative volumes of broad frequency ranges. That's your first mistake. Higher frequencies contain more energy and are going to be perceptually louder, which is why most music has low-frequency boosts to sound neutral. Open up a track in Audacity and run a plot spectrum and see for yourself. Pink noise listening is far more effective because it presents all frequencies with equal energy, and when listening with the IE800, the spectrum is absolutely dominated by the low frequency rumble. 

Secondly, variances due to anatomical differences are nowhere near significant enough to produce that kind of variability unless you have a major deformity, The size, shape, and frequency sensitivity of the ears is an evolutionary adaption to improve survivability. If there was a large degree of variance in hearing, then not all individuals would be well-adapted to survival. However, we're all alive today, so it's obvious that our ancestors were well-adapted, meaning that our hearing is generally going to be within an acceptable range of variability, ie, is not all that different from one another. 

The IE800 are boosted by almost 10 dB in the low frequencies, which is at least twice of what is required for a neutral-sounding IEM. IEMs may require slightly more bass compensation to sound neutral, but if so, it's maybe only ~2 dB or so above full-sized headphones, not +10 dB

Stefraki's picture

They are off for repair right now, so we will see. I should say I have a history of "neutral" headphones (or what has been passed off as neutral in these circles) - I am not coming from a "bass head" fraternity. The iE800 I have/had are not boosted in the bass, they were perceptually even in volume all the way down... but I will run the tests you suggest when they get back. Keep an eye here if you are interested.

Stefraki's picture

Sennheiser eventually sent me back brand new IE800 and they are no different - I ran your tests as well as mine - the bass is perceptually even.

In fact if I compared them to the Focal Spirit Classic - the Classic sound clearly bassier.

T N Args's picture

So you applied the green dotted line instead of the black line? Why didn't you follow the black line in the bass region?

I'm very disappointed that you didn't simply apply the Olive/Harman line, instead of making your personal modifications to it. All just because of 'what you are used to'. That decision of yours did us all a disservice.

Tyll Hertsens's picture

....I used the black line, except for above 10kHz. Above 10k the green line dives very steeply, when I applied the curve the treble above 10kHz would rise even more sharply than the curve shown above---which are still too emphasized acording to many comments above. 

It's somewhat likely the way Harman and I do measurements differ, and it very well could be that differences in calibration account for the measurements above 10kHz. Above 5kHz all these systems have trouble due to ear canal and conch resonances. In fact, this area of research---high frequency measuring of audio with ear canal simulators---is ongoing. Paul Barton, of PSB and NAD, is currently working on a new ear canal simulator up at NRC in Canada to help solve some of these problems.

The most important thing to me was to give a first shot at the new curve and because of the differences in measurement equipment I felt qualified to modify the curve somewhat for my measurement system. To me, the very cool part of the test was the flattening of the resultant compensated curves between 1kHz and 8kHz. This has been a consistant problem area in my measurements. The new curve does a good job of making this region look flat on flat sounding headphones, so I'm happy with that.

The other very interesting concept brought out in these measurements is people's feelings that the new curve represents too much bass, making headphones look bass shy, according to many people's posts here. This one is far more tricky. I do think headphones sound more pleasing with a bass boost, but I too, along with many other posters here, think the Harman curve is too bass boosted. I think 3dB is just right; I think many here would say that's probably a bit too much and 2dB would be better. This subject of how much bass boost is correct is likly to be with us for some time to come.

Stefraki's picture

Not sure you should be too bothered with what others say, that would the the tail wagging the dog.

I went through a period of listening to progressively more bass-shy headphones, and it got to the point where I had a modded-YH1 that sounded painfully bass heavy to me and I couldn't reduce the bass anymore through modding and consigned them to a drawer. After a couple of years of listening primarily through speakers with a sub painstakingly calibrated to flat, I listened to those same headphones a couple of months ago and they were clearly extremely rolled off in that department.

We adjust to frequency response irregularities when we regularly use similar headphones, our minds "normalise" them.

If the research shows a curve is right, go with it. People will have to learn to accept that things aren't as they thought they were. 

DanTheMan's picture

I got it by measuring my in ear response in a calibrated stereo.  Audyssey-style and removing the dip I got.  I'm not exactly sure how to get it from what you've measured.  It definitely works well for me.

DonGateley's picture

Tyll, Is the data you measure and plot available digitally or is that proprietary?

DonGateley's picture

I must assume that not responding to the question means that the data is proprietary and unavailable.

Tyll Hertsens's picture

.....yes, the data is owned by Source Interlink Media and is unavailable for use.  Sorry.

DonGateley's picture

Do you know if any kind of restricted licensing might be considered?  I don't want to display or expose it, just use it for DSP and analysis purposes.

Tyll Hertsens's picture

...I don't think so.  I have on occasion provided limited measurements for university students wirking on a paper, but other than something like that I'm afraid it needs to remain proprietary.

DonGateley's picture

I would appreciate the opportunity to discuss this with you via email.

barney's picture

Hi Tyll,

I apologize for being a bit thick about this, but I'm not sure I understood your original post.

  1. Is the first graph -- the one with the purple line -- the correction you applied to the individual headphone curves? 
  2. Is the black line on the second graph the HK target curve?
  3. Is a "target curve" what an ideal headphone should measure?
  4. Where does the green line on the second graph come from? 
  5. How might these target curves be used to equalize headphones you've measured? If, say, a headphone is 5 dB "down" at 3kHz relative to the target curve, would one equalize it by boosting 3kHz 5 dB?

Thanks! And, once again, apologies for failing to understand what everyone else seems to take for granted.

 

Barney   

Tyll Hertsens's picture

1. Yes.

2. Yes...though preliminary.

3. Yes, at the eardrum.

4. IIRC, it's the curve you'd get at the ear drum in a room with speakers that measure flat...I think.

5. Basically yes. 

barney's picture

Hi Tyll,

I corrected the response of my LCD-3's by using graphic equalization to correct the error implied by your posted response curve. (I also used a parametric equalizer to correct what I asusme to be my particular headphones' quirks.)

My first response: the sound was reminiscent of a very comfortable, very enjoyable stereo system in a good room. I'm not certain it was more accurate than the same 'phones equalized to a curve with a less dramatic dip between 100 and1000 Hz, and a less dramatic high frequency peak, but I'll do more listening.   

Hear The Difference's picture

Barney, thanks for asking the question I was too timid to ask (in effect, "Just what ARE these curves?")  And Tyll, thanks for confirming his suspicions.  NOW I can read your carefully-prepared article and start to understand the work that you report.

Hear The Difference's picture

To clarify my understanding (and to spawn correction if I don't have this right), here is what I think is going on...

1) Original headphone curve (e.g., from innerfidelity.com data sheet)

2) Seeks to approach the Harman target response curve,

3) By applying the purple compensation curve that starts this thread...

4) Resulting in a final headphone curve that comes closer to target (e.g., look how the higher frequency is raised).

Right?

vilhoke's picture

As I see it, the purple curve is the old compensation curve, which has been previously utilized to compensate the raw data (the grey curves), and has been applied in all of the measurement data sheets in the Inner Fidelity database. The black compensation curve has been used (with slight corrections at 10-20 kHz) in the new frequency responses shown in the current post to compensate the same raw data. So the purple and black curve are two competitive compensation curves.

So:

1. Is the FR compensated with the old, purple compensation curve.

2. The black Harman target response curve has been sligthly modified and used as a compensation of new FR curves shown in this post...

3. ...instead of the purple compansation curve.

4. Resulting in a FR curve that should be a flat line in the case of "ideal headphones".

I may have misunderstood something too, but this is how I have interpreted the process.

Hear The Difference's picture

vilhoke, I think I got it wrong and you got it at least more right... thanks.

So... HOW does that purple (or black) compensation curve get "applied" to that raw data?  By arithmetic in a spreadsheet?  By equalization with a multiband equalizer?

 

Thanks!

3amsleep's picture

I own the ATH m50 and the SE 535, and I have to say that the SE 535 sounds spot on. I'm very accustomed to Studio monitors and my EQ for the SE 535 is boost 0.5db at 31hz and 62hz, 1db at 4khz (heard that drop from the first time I've put them on) and 0.5db at 8khz (which is not really necessary but I like it). With that EQ the sound is very similar to a flat studio sound.

I haven't used the m50 for a while, but that curve looks about right, though I think they actually sound more boomy than that.

Awesome article!

X