[ADUHF]

I already saw the the Axpona version, which was an excellent presentation. I don't think anybody has collected that much knowledge to conduct such presentation at the time.

Probably true.

Take away from his analysis is that ideally, the response should roll-off where he's stating, which I'm not quite sure about. It could just be his rig rolling off too quickly, and his view based on his rig in combination of what he has heard (with bright headphones) and compared with measurement. Making correlation based on lots of data he experienced.

Bear in mind that the Harman curve has been revised several times since these videos were made. And the responses in both the bass and treble tweaked to better match the subjective preferences of a few different demographic groups. I believe the younger male listeners preferred more bass and less treble. And the older demos were essentially the reverse, possibly due to hearing loss in the higher frequencies.

I believe the average bass preference for all listeners was revised upwards though to about 6 dBs above the baseline in the lower midrange.

 

[gooeyrich]

 

[ADUHF]

Excellent panel. The discussion raises more questions than answers though.

 

[ADUHF]

Take away from his analysis is that ideally, the response should roll-off where he's stating, which I'm not quite sure about. It could just be his rig rolling off too quickly, and his view based on his rig in combination of what he has heard (with bright headphones) and compared with measurement. Making correlation based on lots of data he experienced.

More to your point (which I seem to have missed in my previous post)... If you're referring to the rolloff in the treble, that actually has more to do with the measuring system than the headphone being measured. The peaks that occur at around 3 kHz on most raw frequency plots are largely the result of the characteristic resonances within the simulated ear. And that is why they show up fairly consistently in the same place, over and over again, on many headphones.

If you take basically any large grouping of headphones, and average all their raw frequency responses together, even if they are not the most neutral headphones, I can pretty much guarantee that you will see a peak showing up close to that same spot. And maybe also one close to 8-10 kHz on many of the measuring systems. Because those are the resonant frequencies of a typical ear canal of about 25 mm in length, used in the measurement systems.

If you change the length of the ear canal, or take the measurements on different individuals with different ear geometry, then the resonances will likely show up at somehat different frequencies. The dummy heads and ears used for these type of measurements are based on the average dimensions of humans though, laid out in various standards. So that's why the resonances are usually pretty similar in the upper mid-range and treble, particularly in the 3 kHz range.

The curves on the plot below were likely made on three different measuring systems, for example. They all have peaks close to 3 kHz in the upper midrange. But it's shifted a little more to the left on the first curve, which is Olive-Welti. And shifted a bit more to the right on the last curve, which is the current Harman target. Those small shifts in the peak frequency are most likely due to minor variations in the ear geometry and resonant frequencies of the HATS systems used for the three measurements.

 

[ADUHF]

I think it's gonna be fun trying out some different equalization curves on my 250-ohm DT-770's btw, to see what sounds best to my ears. There is a fair amount of variation in the recommended EQ curves on AutoEQ, for example. And my initial experiments suggest that I may like a somewhat darker/bassier overall response, and perhaps also more U-shaped response than most of those EQ curves. Or the raw FR plots shown on the graph above. Whether that'll hold up with more listening though, I'm not sure.

I didn't really have the luxury of that sort of info available on my old AKG K553's. Because nobody seemed to wanna do any graphs of those. Good times.


Maybe I should just embrace my basshead-ish tendencies. And just get a Sony or Bose. And call it day. Life might be happier that way.

 

[ADUHF]

I don't know if you ever got a chance to read the articles on the Harman curve that Tyll Hertsens wrote on Inner Fidelity. But at one point he was invited into Harman's reference listening room to take some measurements with his HATS system. And my recollection is that his measurements were a pretty close match to the frequency response of the AKG K371. (AKG is part of Harman, for those who don't know.)

What he measured in their room though (and stated in his articles) was NOT the response of two anechoically flat speakers. Because the speakers in the room were actually electronically tuned to Harman's preferred in-room listening curve.

So it is possible that what you are hearing on the AKG K371 is similar to the "house sound" in Harman's reference listening room.

Metal571 seems to concur with the above here. So it may not be all in my head.



As its description implies, Harman's preferred in-room loudspeaker response is also based to some extent on subjective listening tests. That is my recollection anyway. (Mr. Olive is welcome to chime in though, and correct all the above.)

The Harman curve isn't really "B&K-ish" though, as Metal at first suggests. The original B&K speaker curve that he's referring to is more mid or lower mid-centered. Whereas the Harman curve is more shelved up in the bass/sub-bass (as he later points out).

Based on some of his posts, B. Katz uses (and recommends) a somewhat B&K-like room curve for his studio. If you are mixing and mastering with an in-room response like that though, it will tend to make your mixes more U-shaped. Which may actually be the intent.

This is one of the reasons though why (imho) a studio or mastering suite may not necessarily be the best model for a neutral response. Because the acoustics and response in a studio may deliberately be "tuned" differently than the intended listening space, for engineering or other reasons.

Some of the highest-end Genelec loudspeaker monitors have an almost totally flat direct/anechoic response. And also a very linear in-room response (based on the Harman spinorama) that slopes downward from sub-bass to high treble in almost a straight line, which is quite remarkable. So there isn't really any shelving in their response per se.

 

[ADUHF]

Because those are the resonant frequencies of a typical ear canal of about 25 or 26 cms in length, used in the measurement systems.

Should have been millimeters rather than cms. Corrected above. A 25 centimeter ear canal would be waaaaay long.

We're still sort of adapting to the metric system in the U.S.

 

[ADUHF]

Dr. Olive seems to acknowledge that there are disagreements re the best way to handle the room response in this recent tweet btw...

https://twitter.com/seanolive/status/1331322019800084482

With the exception of the higher-end Genelecs (which are undoubtedly extremely well-designed speakers, but probably not the norm of what's used in both homes or mastering suites), most of the room measurements I've seen so far seem to have a flatter area somewhere in the middle range, a rise in the bass, and a small rolloff in the treble.

And the curves used by auto sound mfrs generally seem to have a similarly serpentine shape. The in-room response of speakers can vary quite a bit though, based on the size of the room, speakers, etc. And in some instances, the rise into the bass may be much more gradual, beginning well up in the midrange. And looking more linear overall. While in others it may be more abrupt, and bit closer to the more blatantly shelved-up response of Harman's curve.

Imo, a curve which is rounded (and also elevated) in the bass, and also rounded off in the treble, and somewhat flatter in the midrange will tend to de-emphasize the mids, giving the sound a bit more "depth" and punch in the bass and upper mids/low treble than a slope with a more linear falloff would. I'm honestly not sure which is "right" though... Or if there really is a "right" answer to this question.

I don't work in music production, but I get the distinct impression from what I've heard and read that most audio and mastering engineers are concerned about achieving a fairly neutral response. But only up to a point. What usually seems to be top-most on their mind is what the client thinks. And how they're going to be listening to the final result, in their homes, car, ear or headphones, etc. Because that is where the tires hit the proverbial road, so to speak. And if their clients are not listening on mostly neutral devices, then there may actually be a disincentive for them to religiously try to follow that model.

If we do eventually all agree on what neutral is though, I would hope it is something close to the perceived sound of a pair of anechoically flat speakers, with decent bass extension and smooth off-axis response, in a typically reflective listening space in a home. Because that is really the best place to listen imo. Perception can be a tricky thing though. And I unfortunately don't know enough about the science of how sound behaves in rooms to really know what may be the best model for that sort of thing. I'm sure there's lots of good info on that topic in Floyd Toole's book though, for those who want to learn and study more about it.

 

[SilverEars]

@ADUHF

I was curious what you think of ear gain humps on both full-sized headphone measurements and iems? Does it seem right to have same ear gain level on iems just like full-sized headphones? Full-sized rig does reflect the pinna gain due to sound interacting with the pinna, which makes sense. On the iems, there is no pinna interaction, but should the iems be tuned for artificial pinna interaction to sound more of what is expected?

 

[ADUHF]

@ADUHF

I was curious what you think of ear gain humps on both full-sized headphone measurements and iems? Does it seem right to have same ear gain level on iems just like full-sized headphones? Full-sized rig does reflect the pinna gain due to sound interacting with the pinna, which makes sense. On the iems, there is no pinna interaction, but should the iems be tuned for artificial pinna interaction to sound more of what is expected?

Well, it's an interesting question. And one that I honestly haven't given too much thought, since I don't use IEMs myself. But I'll try to give my opinion anyway.

If the outer ear or pinna does indeed have some bass-boosting effect on the sound before it enters the inner ear, then theoretically it might make sense to add something comparable in the way of a bass boost to an IEM, to account for the loss of that.

I think there are some important caveats to that though.

When you're listening to speakers or sound in a room (which is presumably still the model we're using for all this stuff), the sound has a number of different interactions with different surfaces and objects before it finally reaches your ear-drum where you hear it. It interacts with the room first, then with your body and head, and then with the outer ear, and finally with the ear canal and parts of the inner ear.

Whenever the subject of target curves and HRTFs for headphones comes up, there is often a tendency by some to simply throw their hands up in anguish, and proclaim that no two people or ears are alike, therefore there can't be just a one-size fits all solution to a neutral response. And we shouldn't even bother trying to find one. And I think there's probably some truth to that. The variations in human anatomy really only begin to come into play though after the sound leaves the speaker and interacts with the room, and reaches your body. From that point on, the differences in the shape of your body, head and ears can potentially have a different effect on the exact frequencies that you hear at the DRP or ear-drum reference point, versus someone with a different anatomy.

Basically, the further away you get from the sound source (a speaker in this case), and closer you get to the DRP, in terms of the chain of acoustical interactions, the greater the effect that differences in human anatomy can have on the physically measurable characteristics of the sound that you hear at the ear drum. And the more difficult it is to accurately model or predict what the sound is like at the point of measurement.

...Which means that it is harder to accurately model the sound in an IEM (which has no interaction with the outer ear), than it is to model the sound in an on-ear headphone (which has some degree of interaction with the outer ear), or an over-the-ear headphone (which has an even greater degree of interaction with the outer ear). Because in an IEM, the physical effects of the outer ear geometry (and any variation there is in that) are simply missing. Which in turn can lead to greater variations in what is measured and perceived as "neutral" at the DRP.

 

[ADUHF]

There are several conclusions that one might potentially draw from the above, if you accept the general premise laid out there about the chain of acoustic interactions, and its effect on the perceived sound at the eardrum. One is that personalized HRTFs are probably more important with IEMs than with other kinds of headphones, because there's more of an individual's ear anatomy missing in the sound of an IEM than in other types of headphones.

Another conclusion you could draw is that a good over-the-ear type of headphone would probably stand a better chance of accurately modeling and reproducing the neutral sound of a speaker in a room for a larger percentage of the population than an IEM would. Because the over-ear HP takes more of the physical variations in the shape of the outer ear and the potentially different effects it can have on the measured sound into account than an IEM would.

Another conclusion you could draw is that it's probably a little easier (but also still rather hard, in any case) to find where a neutral target is for a broader segment of the population with an over-the-ear headphone than it is with an IEM. There can be some variations in the measured sound of over-the-ear headphones for other reasons though, such as fit, seal, etc..

 

[SilverEars]

Well, it's an interesting question. And one that I honestly haven't given too much thought, since I don't use IEMs myself. But I'll try to give my opinion anyway.

If the outer ear or pinna does indeed have some bass-boosting effect on the sound before it enters the inner ear, then theoretically it might make sense to add something comparable in the way of a bass boost to an IEM, to account for the loss of that.

I think there are some important caveats to that though.

When you're listening to speakers or sound in a room (which is presumably still the model we're using for all this stuff), the sound has a number of different interactions with different surfaces and objects before it finally reaches your ear-drum where you hear it. It interacts with the room first, then with your body and head, and then with the outer ear, and finally with the ear canal and parts of the inner ear.

Whenever the subject of target curves and HRTFs for headphones comes up, there is often a tendency by some to simply throw their hands up in anguish, and proclaim that no two people or ears are alike, therefore there can't be just a one-size fits all solution to a neutral response. And we shouldn't even bother trying to find one. And I think there's probably some truth to that. The variations in human anatomy really only begin to come into play though after the sound leaves the speaker and interacts with the room, and reaches your body. From that point on, the differences in the shape of your body, head and ears can potentially have a different effect on the exact frequencies that you hear at the DRP or ear-drum reference point, versus someone with a different anatomy.

Basically, the further away you get from the sound source (a speaker in this case), and closer you get to the DRP, in terms of the chain of acoustical interactions, the greater the effect that differences in human anatomy can have on the physically measurable characteristics of the sound that you hear at the ear drum. And the more difficult it is to accurately model or predict what the sound is like at the point of measurement.

...Which means that it is harder to accurately model the sound in an IEM (which has no interaction with the outer ear), than it is to model the sound in an on-ear headphone (which has some degree of interaction with the outer ear), or an over-the-ear headphone (which has an even greater degree of interaction with the outer ear). Because in an IEM, the physical effects of the outer ear geometry (and any variation there is in that) are simply missing. Which in turn can lead to greater variations in what is measured and perceived as "neutral" at the DRP.

The ear gain is effecting the upper-mids, not the bass, so there shouldn't be bass variations based on pinna gain. When there is no effect of head, torso, pinna, for iems, only variations is the ear canal itself. In the case of circumaural, there is variations of pinna geometry plus ear canal, so therefore, I would expect to get better consistancy in an statistical avg target response with iems due to the less variables. And with speakers, more variables of head, torso, pinna, and canal, which creates more complexity and increases subjectivity.

 

[ADUHF]

The ear gain is effecting the upper-mids, not the bass, so there shouldn't be bass variations based on pinna gain.

Interesting. I didn't know that.

As I said above, I haven't really given the sound of IEMs a whole lot of thought. I think the above principals should still apply though, regardless of the frequencies effected.

The graphs I've seen of IEMs often look more boosted in the bass region though. And the Harman in-ear target also seems to show that adjustment. So that's what I assumed you were referring to in your original question about IEMs above...

If the bass boost on the IEM target isn't related to outer-ear geometry and it's effects at the DRP though, then I guess it's being done for some other reason.

I see what appears to be a slight gain in the upper-mids/low-treble on the above IEM curve as well. Is that what you're referring to?

 

[SilverEars]

Interesting. I didn't know that.

As I said above, I haven't really given the sound of IEMs a whole lot of thought. I think the above principals should still apply though, regardless of the frequencies effected.

The graphs I've seen of IEMs often look more boosted in the bass region though. And the Harman in-ear target also seems to show that adjustment. So that's what I assumed you were referring to in your original question about IEMs above...

If the bass boost on the IEM target isn't related to outer-ear geometry and it's effects at the DRP though, then I guess it's being done for some other reason.

I see what appears to be a slight gain in the upper-mids/low-treble on the above IEM curve as well. Is that what you're referring to?

All these responses including targets are merely tonal responses, but imaging is what I believes differentiates the 3 types. Speakers and circumaurals have pinna interactions which causes sound directional information, and iems do not have such information due to strictly ear canal interactions.

 

[ADUHF]

Frequency response in relation to imaging is not something I've given much thought. So feel free to elaborate further on that, if you wish.

I think of stereo imaging as having more to do with clarity and the symmetry of the drivers though. So a headphone with good imaging to me would be one with lower distortion, good clarity, and good balance between the sound of the left and right drivers across the frequency spectrum.

Some of these audiophile terms like imaging, sound stage, transient response, etc. elude me a bit though.