[Àedhàn Cassiel]

I suspect that something about the driver size/angle is making a difference here. My ear lays relatively flat against the inside of the LCD-4... It lays relatively flat against the Vérité Closed. The Atticus has much deeper cups that my ears don't touch at all. The HD800 does too... but the HD800 angles facing the ear, while the Atticus still aims perpendicular to the head. So the Vérité Closed sounds remarkably close, but my suspicion is the Atticus has a less predictable number of interactions with my ear. Specifically, it sounds too bright to make a convincing LCD-4 impression. So maybe there's more pinna gain amplifying the upper midrange when it hits my ear from the side like that. Another theory could be that the 14 kHz peak is narrow enough that the AutoEQ doesn't correct it properly, to test that I should try increasing the maximum gain on the generated EQ. It's kept relatively low by default to allow it to be used without tons of extra headroom.

Purr1n is a smart guy I respect. I'm open to the possibility, but not yet convinced CSD shows something separate from frequency response. I'm not even sure what explanation he would give for what's going on acoustically if he thinks it truly is something separate, as in some contexts if I'm interpreting right I've heard him admit that it is just a different way of looking at FR (albeit one that could be useful). Even if CSD could show differences between two headphones tuned exactly the same, it's not clear how audible by human beings these remaining differences would be once FR was perfectly equalized. The couple studies that have been done so far show that frequency response accounts for >85% of preference ratings for headphones, and it's possible the 15% gap was due to imperfections in the virtualization process.

 

[dazzerfong]

I suspect that something about the driver size/angle is making a difference here. My ear lays relatively flat against the inside of the LCD-4... It lays relatively flat against the Vérité Closed. The Atticus has much deeper cups that my ears don't touch at all. The HD800 does too... but the HD800 angles facing the ear, while the Atticus still aims perpendicular to the head. So the Vérité Closed sounds remarkably close, but my suspicion is the Atticus has a less predictable number of interactions with my ear. Specifically, it sounds too bright to make a convincing LCD-4 impression. So maybe there's more pinna gain amplifying the upper midrange when it hits my ear from the side like that. Another theory could be that the 14 kHz peak is narrow enough that the AutoEQ doesn't correct it properly, to test that I should try increasing the maximum gain on the generated EQ. It's kept relatively low by default to allow it to be used without tons of extra headroom.

Purr1n is a smart guy I respect. I'm open to the possibility, but not yet convinced CSD shows something separate from frequency response. I'm not even sure what explanation he would give for what's going on acoustically if he thinks it truly is something separate, as in some contexts if I'm interpreting right I've heard him admit that it is just a different way of looking at FR (albeit one that could be useful). Even if CSD could show differences between two headphones tuned exactly the same, it's not clear how audible by human beings these remaining differences would be once FR was perfectly equalized. The couple studies that have been done so far show that frequency response accounts for >85% of preference ratings for headphones, and it's possible the 15% gap was due to imperfections in the virtualization process.

CSD is simply an extension of a FR. Think of it like FR over time. It explains how some peaks come to be: some peaks are because the sound 'lingers' over a longer period of time due to poor damping, and others simply are because the initial response is large.

Overall, FR is still the predominant factor, but CSDs reveal more than a simple FR may suggest.

 

[Redcarmoose]

Look at the Sony R-10, a $5000 headphones which did midrange really good and had a grand following. Even today they are regarded as a pinnacle of the art of headphone making.

I’ve heard the R-10 on an optimal system. Yet years go by and a new younger and different set of engineers make the new Sony sound. Now we have MDR-Z7 and MDR-Z7#2. We have the MDR-Z1R.

It seems to me the R-10 Engineering Team would have completely thought the MDR-Z1R was wrong sounding. There is a physical side to the two headphones which has me guess that an R-10 could not be EQed to sound exactly like the MDR-Z1R. If anything the two headphones are polar opposite yet both Sony flagships in their day and age. There is the group that hate the modern advancements. There is a group that feel the MDR-Z1R is not technically superior to the R-10.

With this single narrative I conclude that headphones probably are not simply differentiated by distortion and sound signature. But that even at the end if it was true it wouldn’t matter. People have their own ideas of right and wrong. The idea of flat neutral is also up for argument. Also my gut tells me this headphone personally thing is going to be much like everything else in the audio world. Everything simply has it’s own personality. You either enjoy that personality or it sounds off. There is such a wide range of soundsignature responses that it’s almost infinite. Then their is this instrument timbre and character which while it can be emulated often by different headphones is never exactly the same. Look at the effect of the R-10 wood on resonance and compare that to the paper and plastic, plus metal mesh configuration of the MDR-Z1R.

No matter what electronic tom-foolery is accomplished upstream those resonate characters are going to be vastly different. Sound is so delicate at times just a slightly different style of wood on an acoustic guitar top completely rearranges the personality at hand.

 

[castleofargh]

Who would've thought headphones are not only differentiated by sound signature and distortion.................

Unfortunately, there's still a lot to transducer tech that we don't fully grasp. Analysing CSDs, however, is starting to become more of a thing as it provides a more detailed look at what sound is actually doing.

You're not wrong, but I'd argue that most of the divergences we notice are in effect still mostly signature and distortions. Simply because in most cases we do it wrong(often we don't really have a choice). Taking a headphone response of some headphone measured online instead of our pair. Then applying an EQ without thinking too much about it, to get a result that in the best scenario will give the same signature as found with a specific placement(maybe not our own) on a dummy head. That gives zero guaranty that we're getting the same signature at our eardrum. In fact it's almost certain that we're not.

Then there is always the possibility that we're boosting a frequency that the headphone will not be able to boost with the asked magnitude at our listening level. And instead we'll get wrong response and most likely, loads of distortions.

Another problem is that most people try to turn a donkey into a race horse. That process would obviously have more chances to succeed the other way around. Using a headphone with smooth response and low distortions and apply EQ to it.


My long point is that the same signature already does a great deal subjectively if we actually get to hear the same signature. And if we add a little more by capturing an impulse response instead of only the frequency response. And use convolution after that, then I'd say that the average audiophile would be surprised by how strongly he underestimated what could be done. That's still not complete, we only have the linear stuff with the impulse, but it's already pretty cool IMO.

 

[dazzerfong]

You're not wrong, but I'd argue that most of the divergences we notice are in effect still mostly signature and distortions. Simply because in most cases we do it wrong(often we don't really have a choice). Taking a headphone response of some headphone measured online instead of our pair. Then applying an EQ without thinking too much about it, to get a result that in the best scenario will give the same signature as found with a specific placement(maybe not our own) on a dummy head. That gives zero guaranty that we're getting the same signature at our eardrum. In fact it's almost certain that we're not.

Then there is always the possibility that we're boosting a frequency that the headphone will not be able to boost with the asked magnitude at our listening level. And instead we'll get wrong response and most likely, loads of distortions.

Another problem is that most people try to turn a donkey into a race horse. That process would obviously have more chances to succeed the other way around. Using a headphone with smooth response and low distortions and apply EQ to it.


My long point is that the same signature already does a great deal subjectively if we actually get to hear the same signature. And if we add a little more by capturing an impulse response instead of only the frequency response. And use convolution after that, then I'd say that the average audiophile would be surprised by how strongly he underestimated what could be done. That's still not complete, we only have the linear stuff with the impulse, but it's already pretty cool IMO.

Fair point.

I think this topic would be even more interesting if we could have a way of estimating what everyone's optimal frequency response would be, considering that the main reason there's so much subjectivity is because everyone's preferences are different.

 

[bigshot]

FR preference is a range. The Harman curve for headphones is a good generalized target. From there, each individual can stray a bit here and their to suit their own ears and tastes, but it isn't likely to be a huge deviation.

 

[Àedhàn Cassiel]

And use convolution after that, then I'd say that the average audiophile would be surprised by how strongly he underestimated what could be done.

I think AutoEQ is already using it by default, but would you mind explaining what convolution is for me?

FR preference is a range. The Harman curve for headphones is a good generalized target. From there, each individual can stray a bit here and their to suit their own ears and tastes, but it isn't likely to be a huge deviation.

Personally, I can honestly say that I literally hate the Harman curve. I even tried going in fresh several days with it turned on before I even put the headphones on and adjusted the volume. I didn't get used to it and start liking it more. Well, having nothing to immediately compare it to made me hate it less instantly than listening otherwise and then turning it on and readjusting volume. But I never enjoyed it on anything I own. And I was always happy the moment I cut it off or at least reduce it. All I found it useful for was a convenient way to reduce peaks when they bothered me (not using the full values, or all the values).

The most significant place my taste differs is I've found I prefer the midrange to start rising earlier, perhaps even slightly in the low mids, hump around 1 to 1.5 kHz, and then dip fairly sharply (almost 10dB) at 2kHz while not rising nearly as much for the 3kHz peak. Other target curves add less 3kHz peak than Harman, and I prefer even less than those. I've figured out that this is the biggest reason I loved the HD800 so early on - and also the Audeze LCD line. One's extremely bright and one's extremely dark... But they both have the tendency to go this way in the mids.

That upper treble rolloff is also viscerally absurd to me. Even the LCD-4, which everyone knows as "dark," has more upper treble.

 

[bigshot]

That is why I say that the Harman curve is just a starting point. It's a generalized setting designed to hit the middle of the bell curve. But if you are personally at the outside edges of the bell, you will need to adjust. The Harman curve comes very close for me. I generally just need a little adjustment in the upper mids, because my hearing is a little bit sensitive there and I need to pull it back a few dB to make listening comfortable.

 

[castleofargh]

I think AutoEQ is already using it by default, but would you mind explaining what convolution is for me?

Convolution is math voodoo. For audio application, the most telling application is a reverb plugin. You go to some place, record a clap, bang, anything impulse like. So the recorded impulse will also have how the room echoed that clap, it's not just the frequency response, but also how the sound in the room decays over time. With convolution you can run any audio signal as if it was played in that room with the added change in response and the reverb(that's the voodoo part IMO, even magnets aren't that amazing).

From headphone to headphone the "rooms" are too small to have much of an audible impact beyond frequency response, so we can expect good result just using the FR(if it's the right one!!!!, more on that later), but let's say that one headphone has a particularly bad damping at some frequencies and the decay is so long that we might notice an impact. If we captured an impulse(or a sine sweep), we could get the sound to have that long decay baked into it when using the other headphone. It might not be useful or even desired, but it's one little step closer to getting the same sound. autoeq might use an impulse and convolution, but the information in the impulse is only the frequency response picked from some online graph, it has no information in the time domain(not even phase shift for the headphone). not that it can't, it's just that what's available everywhere are FR graphs so that's what is used.

and that leads to me mentioning the right frequency response, and by that I mean trying to record the sound with mics in our own ears. That way you don't just get "a" frequency response, you get one from wearing your headphone the way you're used to, and the frequency response measured is the one near the entrance of your ear canal as impacted by your own outer ear. Which is likely to be closer to your own experience of the headphone than the response captured by some dummy head or cardboard box with a microphone in it.

@jaakkopasanen makes complete use of impulses for his other project to simulate speakers with headphones. There you get delays between speakers, reverb, frequency response. I hate the "more is better" argument, but in this case, it's often true for the listener ^_^. You can use the same idea for headphone to headphone(but once you have all the tools for speaker simulation, you might not care much about imitating another headphone anymore).

 

[Àedhàn Cassiel]

Yeah, I've been planning to check out his other projects next. But I've tended to prefer headphones to speakers so far up until now, and I don't think I can understate just what a revelation this LCD-4 emulation is for me. You know how when they're talking about piracy they say, "You wouldn't download a car?" I feel like I just downloaded a car.

 

[Blackwoof]

Exactly. If a set of cans can play a wide range of frequencies loud, it can be EQed into anything you want. The only exception is sub bass, which depends a lot on kinesthetic vibrations. But no headphone can do that like a subwoofer can.

You can easily crossover a woofer with a headphone/IEM, I've read many doing that with their HD650's. The subpac is mini woofer that look like a vest that gives that sub bass punch that headphones can't do.

Which I'm thinking of doing since i don't even expect the 1 person apartments in our town, To be fit for a loudspeaker set up. Assuming i don't just go 2.1 studio monitors near fields with the ER3XR as my back up.

 

[bigshot]

In the 70s they had a gizmo called Bone Fone like that

 

[BrownBear]

I don't understand why a person would want a bunch of different sounding headphones. I have a specific target I'm looking for. Once I reach that, almost all my music sounds great and I'm done... ready to listen to music and stop fussing with the equipment. I think people who want to have a bunch of different kinds of sounds just haven't sat down and analyzed sound to determine their own target response. Instead they listen to one coloration until they get tired of it and then move on to another. And on and on back and forth. If they'd just expend the energy to define their goal, they wouldn't be see sawing so much. They'd arrive at their happy place and not have to fuss any more.

This is a great post and something to think about. I do have several headphones myself, however I find myself always reaching for the same pair. All my music sounds good with them and they are pleasing to my ears, so basically, that's all I need.