[JohnYang1997]

There was one document from b&k that showed the peak at 13.5khz is very high at 35+db with real human ears. That's the reason why the highest quality 711 couplers all have that high amplitude peak.

 

[csglinux]

There was one document from b&k that showed the peak at 13.5khz is very high at 35+db with real human ears. That's the reason why the highest quality 711 couplers all have that high amplitude peak.

That could be an interesting read. Do you have a link for that? Jude had actually mentioned to us that B&K are working on their own hi-res coupler, based on scans of real ears.

 

[yuriv]

Nice posts @yuriv! Maybe even too nice, because now you'll have intimidated and frightened away potential future contributors to this tour We've not yet had anybody brave enough to volunteer to measure these headphones with a vinyl tube coupler. It still think it would be interesting to see what the responses look like from a similar-length, flat transfer-impedance cylinder.

Aha! Apple USB-C dongle - got it! Yep, that makes sense. I even have one of those lying around in a drawer somewhere. I should test it with REW...

It looks like we're all in agreement on the hi-res (RA0401/2) coupler. What a shame. After seven pages of this thread, we've had zero arguments. Come on guys, this is headfi! Ok, let me put my GRAS hat on for a minute and play Devil's advocate in defense of the hi-res coupler...

...

Zero arguments, because you took off the devil's advocate hat too soon. Lol. I suppose it's better than trolling, or if you prefer, playing the role of provocateur. I'm just about ready to send the package to Europe after finally giving up on figuring out how to measure impedance vs. frequency using the dongles. For this, I had to return to a proper recording interface. That was the last thing on my list of things to measure, and I've uploaded the files to the shared drive. But I you really want to see measurements from a vinyl tube coupler, I can do that. The package will be delayed a few more days, though.

The graphs I'm getting from the Apple USB-C dongle look pretty much like the ones made with a Focusrite Forte, and a MOTU and a Behringer I have around here. I've also been playing with the new Samsung USB-C dongle. It works well too.

Etymotic ER2SE with ER38-15SM tips, measurements made with Apple and Samsung USB-C adapters.​

It looks like the one of the graphs you made using the RA0045. I didn't put a gap in the stem with the ER38-15SM tips, so I also get the huge null. I offset the traces so you can see more clearly that there are, indeed, two measurements.

 

[JohnYang1997]

That could be an interesting read. Do you have a link for that? Jude had actually mentioned to us that B&K are working on their own hi-res coupler, based on scans of real ears.

I thought that one from b&k was already available. There was a white paper of how they designed and everything. And I think I've seen Sean Olive already measuring in ears with it. I assume it's going to be much better than the gras one. But not so sure.
And the b&k document for ear impedance is here:https://www.bksv.com/doc/bn0221.pdf
I think many have read this.

 

[csglinux]

But I you really want to see measurements from a vinyl tube coupler, I can do that. The package will be delayed a few more days, though.

If you were willing to do that for us I'd say go for it! It would be an interesting comparison and I don't think the folks in Europe would mind a delay of a few more days. Many thanks for the offer

Thanks to @JohnYang1997 for the link to the B&K paper. I think it might be worth transferring the pm John and I were having over to this thread. John was concerned about the problems with using existing couplers for measuring shallow-insertion IEMs. On that subject, I'd be curious to know if anybody else out there had looked at the GRAS anthropometric pinna? They're allegedly hyper-realistic, with all the typical canal bends and external ear components. Essential for over-ear, probably very helpful for ear-buds and probably somewhat helpful for shallow-insertion IEMs where more of the ear canal is exposed. The downsides (which I guess would apply to any existing or future hi-res couplers too) is 1) cost and 2) it's not a widely-shared standard, i.e., even if it did get closer to the average response of a human ear, it wouldn't be much use for comparing against anybody else's measurements if nobody else in the world has such a device. I guess somebody had to be the first person to blaze the trail(?!). In the meantime, one of the advantages of us all having (roughly) the same coupler is that, having gained a bit of confidence in measuring a couple of common IEMs, we could potentially share future measurements and might actually be able to make meaningful comparisons (at least in terms of deltas) across rigs.

 

[JohnYang1997]

If you were willing to do that for us I'd say go for it! It would be an interesting comparison and I don't think the folks in Europe would mind a delay of a few more days. Many thanks for the offer

Thanks to @JohnYang1997 for the link to the B&K paper. I think it might be worth transferring the pm John and I were having over to this thread. John was concerned about the problems with using existing couplers for measuring shallow-insertion IEMs. On that subject, I'd be curious to know if anybody else out there had looked at the GRAS anthropometric pinna? They're allegedly hyper-realistic, with all the typical canal bends and external ear components. Essential for over-ear, probably very helpful for ear-buds and probably somewhat helpful for shallow-insertion IEMs where more of the ear canal is exposed. The downsides (which I guess would apply to any existing or future hi-res couplers too) is 1) cost and 2) it's not a widely-shared standard, i.e., even if it did get closer to the average response of a human ear, it wouldn't be much use for comparing against anybody else's measurements if nobody else in the world has such a device. I guess somebody had to be the first person to blaze the trail(?!). In the meantime, one of the advantages of us all having (roughly) the same coupler is that, having gained a bit of confidence in measuring a couple of common IEMs, we could potentially share future measurements and might actually be able to make meaningful comparisons (at least in terms of deltas) across rigs.

Yes it looks really realistic. And I think it's a good start. However it's very concerning that it's designed to be paired with new Hires coupler first of all and more importantly there isn't a transfer characteristic graph any where. Like old coupler, it doesn't need to be the same physical dimensions (or more likely shouldn't be) but acoustic volume to have the preferred impedance. It just needs to be fine tuned for ear entrance point referred acoustic impedance. It shouldn't be very difficult (meaning very possible just needs a lot to do)to do it. So I would love to see measurements done with the BK hires head. Tho I guess it's still only accurate for deep inserting type.

One thing to point out for different researches. People tend to like " averaged " measured response for everything. Imo median is more appropriate. A good plausible example is, like in the 711 acoustic impedance, it's specified that there should be a peak around 13.5khz +- something, which is very good. Individual human being has a large peak at that region. If we use averaging, we will see a hump near that region which is technically not wrong but very unrealistic. And I think that's what happened with the Harman Target IE. Just something I found and just pop up my head when seeing the word average.

 

[castleofargh]

If you were willing to do that for us I'd say go for it! It would be an interesting comparison and I don't think the folks in Europe would mind a delay of a few more days. Many thanks for the offer

Thanks to @JohnYang1997 for the link to the B&K paper. I think it might be worth transferring the pm John and I were having over to this thread. John was concerned about the problems with using existing couplers for measuring shallow-insertion IEMs. On that subject, I'd be curious to know if anybody else out there had looked at the GRAS anthropometric pinna? They're allegedly hyper-realistic, with all the typical canal bends and external ear components. Essential for over-ear, probably very helpful for ear-buds and probably somewhat helpful for shallow-insertion IEMs where more of the ear canal is exposed. The downsides (which I guess would apply to any existing or future hi-res couplers too) is 1) cost and 2) it's not a widely-shared standard, i.e., even if it did get closer to the average response of a human ear, it wouldn't be much use for comparing against anybody else's measurements if nobody else in the world has such a device. I guess somebody had to be the first person to blaze the trail(?!). In the meantime, one of the advantages of us all having (roughly) the same coupler is that, having gained a bit of confidence in measuring a couple of common IEMs, we could potentially share future measurements and might actually be able to make meaningful comparisons (at least in terms of deltas) across rigs.

IMO it's always the same story, whatever reference we pick is going to be arbitrary and can only ever be treated as such. I can bet an entire potato that my ears aren't like whatever model of pinna, ear canal, or fancy coupler, and I know I will win because my left ear isn't even like the right one, and my ear canals reach all the way down to Narnia.
for all those tools, my concerns are repeatability, and that they facilitate communication/research. if it's a super fancy super accurate stuff but everybody has a decade of habits interpreting the 711 stuff and does it well, while only 10 places around the world will get the new stuff, then there is a legitimate reason to wonder if it's worth it for everybody to try and adapt to that new stuff as a standard. I'm very glad that such products are being made because better human models will improve music for us in the long run. me being a tiny bit obsessed with ultra accurate HRTF modeling, I certainly welcome those new products. but for us here to look at RAW FR graphs and learn stuff about an IEM, I'm gonna say, who cares? the effective variations from user to user, from one insertion to the next, from even the left side of an IEM to the right, are still going to be there in the end and force us to consider a range of tolerance in our graphs that's often bigger or at frequencies more important than the stuff they work on improving. so... good for them, and one day for us all hopefully. but right now those stuff are like helicopters, really cool, but I don't have one.

 

[JohnYang1997]

IMO it's always the same story, whatever reference we pick is going to be arbitrary and can only ever be treated as such. I can bet an entire potato that my ears aren't like whatever model of pinna, ear canal, or fancy coupler, and I know I will win because my left ear isn't even like the right one, and my ear canals reach all the way down to Narnia.
for all those tools, my concerns are repeatability, and that they facilitate communication/research. if it's a super fancy super accurate stuff but everybody has a decade of habits interpreting the 711 stuff and does it well, while only 10 places around the world will get the new stuff, then there is a legitimate reason to wonder if it's worth it for everybody to try and adapt to that new stuff as a standard. I'm very glad that such products are being made because better human models will improve music for us in the long run. me being a tiny bit obsessed with ultra accurate HRTF modeling, I certainly welcome those new products. but for us here to look at RAW FR graphs and learn stuff about an IEM, I'm gonna say, who cares? the effective variations from user to user, from one insertion to the next, from even the left side of an IEM to the right, are still going to be there in the end and force us to consider a range of tolerance in our graphs that's often bigger or at frequencies more important than the stuff they work on improving. so... good for them, and one day for us all hopefully. but right now those stuff are like helicopters, really cool, but I don't have one.

The thing gets even worse when the new stuff is actually worse. If done properly the intuition lf interpreting 711 will still can be maintained and it will be even easier to interpret. Because then you can interpret as is instead of throwing off informations and adding guesses. Perhaps for some/many people who only measure/interpret measurements from 711 but not using ears to actually learn through the measurements, it will be harder. The lost bit is that some of the things learned can be wrong. So if this is what you meant. Then I agree.

 

[yuriv]

This information might be of interest to some of us: https://www.listeninc.com/prediction-of-listener-preference-of-in-ear-headphones-harman-model/

Source: https://www.listeninc.com/wp/media/seq_note_Harman_Predictive_Model_In_Ear_Headphones.pdf​

Yes, it does say Harman IE Target 2019 v2 in the middle pane. The new target curve looks smoother than the 2017 versions, maybe from averaging more results. Eventually I'd like to get a version that can be imported into REW. The SoundCheck software looks interesting too, but it looks like REW covers almost all of its features. If the output shown below really comes from a spreadsheet created by the software, it would be interesting to see how they're calculating the predicted preference:

Same source​

 

[csglinux]

Thanks for sharing this @yuriv Interesting little setup there. Am I right in thinking you'd need two 711 couplers for this experiment?

I'd worry a bit that having perfectly nulled out those peaks and troughs on your coupler(s), that this would transfer across to your ears. It's a cool idea, but probably overly-optimistic in practice. I suspect it would work best on a headphone that doesn't have too many wild peaks and troughs to begin with.

I've got a few doubts about the Harman target itself and I had some recent communications with Sean Olive about this. Firstly, I should concede this could be an irrelevant rant from somebody who doesn't sit squarely in the middle of the bell curve. The Harman target could still be a correct average. But one thing that's always bothered me is the assumption that it's a fixed curve, even for that "average" individual, because of this:

https://en.wikipedia.org/wiki/Equal-loudness_contour

I asked Sean about this, and here was his reply:

"Yes, it is probably true that at lower levels people might prefer more bass to compensate for reduced sensitivity to bass. Our listening tests were done at normal/comfortable levels (average 80 dB (B-weighted) so it should be more representative of typical playback levels than low levels."

There's no other physical reason for that increase in the low frequencies in the Harman target curve (it's not missing as a result of the ear canals being occluded). Sean didn't mention treble, but I think that's an enigma too, because the same effect ought to apply to the upper frequencies at low volumes, and there's no physical reason that I can see for wanting a roll-off beyond 10 kHz:

...unless, maybe(?) you're trying to mimic sound over large distances (maybe large hall or open-air concert) where you have the usual 1/r^2 decay of all acoustic waves but also faster decay of higher-frequency waves due to viscosity:

https://en.wikipedia.org/wiki/Stokes's_law_of_sound_attenuation

But that would seem to depend on what source you're trying to reproduce, e.g., the large concert venue vs. small studio acoustics. I'd rather that were taken care of in the recording, i.e., place the mics wherever you want and then do your best to faithfully reproduce that. Personally, I'd also like to take a little more off the lower treble of that Harman target curve (it looks like they at least started in that direction in 2019):

If anybody wants the data files for these curves, you can grab them from these links (you can then import them as an FR into REW):

https://www.dropbox.com/s/bnleuzf25m575mu/Harman_target_2018.dat?dl=0
https://www.dropbox.com/s/jr6qjz5v45swlib/Harman_target_2019.dat?dl=0

BTW @yuriv, as always, no pressure here, but did you make any progress with measuring our little set of headphones in that simple vinyl-tube coupler?

 

[castleofargh]

Thanks for sharing this @yuriv Interesting little setup there. Am I right in thinking you'd need two 711 couplers for this experiment?

I'd worry a bit that having perfectly nulled out those peaks and troughs on your coupler(s), that this would transfer across to your ears. It's a cool idea, but probably overly-optimistic in practice. I suspect it would work best on a headphone that doesn't have too many wild peaks and troughs to begin with.

I've got a few doubts about the Harman target itself and I had some recent communications with Sean Olive about this. Firstly, I should concede this could be an irrelevant rant from somebody who doesn't sit squarely in the middle of the bell curve. The Harman target could still be a correct average. But one thing that's always bothered me is the assumption that it's a fixed curve, even for that "average" individual, because of this:

https://en.wikipedia.org/wiki/Equal-loudness_contour

I asked Sean about this, and here was his reply:

"Yes, it is probably true that at lower levels people might prefer more bass to compensate for reduced sensitivity to bass. Our listening tests were done at normal/comfortable levels (average 80 dB (B-weighted) so it should be more representative of typical playback levels than low levels."

There's no other physical reason for that increase in the low frequencies in the Harman target curve (it's not missing as a result of the ear canals being occluded). Sean didn't mention treble, but I think that's an enigma too, because the same effect ought to apply to the upper frequencies at low volumes, and there's no physical reason that I can see for wanting a roll-off beyond 10 kHz:



...unless, maybe(?) you're trying to mimic sound over large distances (maybe large hall or open-air concert) where you have the usual 1/r^2 decay of all acoustic waves but also faster decay of higher-frequency waves due to viscosity:

https://en.wikipedia.org/wiki/Stokes's_law_of_sound_attenuation

But that would seem to depend on what source you're trying to reproduce, e.g., the large concert venue vs. small studio acoustics. I'd rather that were taken care of in the recording, i.e., place the mics wherever you want and then do your best to faithfully reproduce that. Personally, I'd also like to take a little more off the lower treble of that Harman target curve (it looks like they at least started in that direction in 2019):



If anybody wants the data files for these curves, you can grab them from these links (you can then import them as an FR into REW):

https://www.dropbox.com/s/bnleuzf25m575mu/Harman_target_2018.dat?dl=0
https://www.dropbox.com/s/jr6qjz5v45swlib/Harman_target_2019.dat?dl=0

BTW @yuriv, as always, no pressure here, but did you make any progress with measuring our little set of headphones in that simple vinyl-tube coupler?

I think you're trying to force the objective approach into a preference curve. if we wished for a sound only, perceived at the ears reference, we've had some for a long time and they always ended up somewhere around what etymotic aimed at.
for the low end:
yes to loudness having an impact, they logically did one graph for one target level. beyond that, it's just a fact that most people, and for once, me included, prefer more subs on IEMs than on headphones where we tend to want more than on speakers(where we kind of expect a solid roll off). why? your guess is as good as mine. some idea of +6 to +10dB in the subs on IEMs has been around for a long time and never really had a sure objective explanation. the lack of tactile bass is probably one aspect? maybe the IEM isolating poorly the low end means we wish to cover that more to get our bass? maybe it's about typical environment? I use IEMs while walking or on a train/plane/bus. the amount of noise is usually significant and the low end noise is the one less attenuated by the IEMs. maybe most people get to ear their heart bit and want the music to cover that? I really don't know why, but I do believe that the average guy(elite audiophools like us might not be that) do not wish for a flat low end on IEMs unless someone confuses him into believing that a straight line on a graph is "neutral".
for trebles:
did they even try to create a trebles curve above 10 or 12kHz? usually between the uncertainty of the measurements and the uncertainty of what will happen in the listener's ear, I expect people to leave that area alone or attenuate it to avoid maybe damaging/unpleasant peaks. when I see those curves with a smooth treble roll off, I can't help but imagine someone applying a low pass filter on account off music content not having much to say up there, and people being less and less sensitive. kind of the "I fixed it boss", that isn't dumb, but also not the actual fix at all.

and for both, we have to keep in mind that it's harder and harder to get a precise idea of the level by ear. I have no issue noticing 0.5dB EQ in the midrange(with repeated instant switch, otherwise I can't tell). but out of 3 or 4 runs of trying to EQ the same IEM by ear with a sweep and a track, it's not that rare for me to end up with more than 4dB difference in the subs and/or trebles. I'm sure the trained guys at Harman can be more consistent, but I wouldn't bet on that for the "random" listeners used for many of the experiments leading to those results.


ps: for dropbox links, replace dl=0 by dl=1

I wouldn't mind a link to the relevant mic and coupler for free too. they logically should come as a package ^_^.

 

[yuriv]

Thanks for sharing this @yuriv Interesting little setup there. Am I right in thinking you'd need two 711 couplers for this experiment?

I'd worry a bit that having perfectly nulled out those peaks and troughs on your coupler(s), that this would transfer across to your ears. It's a cool idea, but probably overly-optimistic in practice. I suspect it would work best on a headphone that doesn't have too many wild peaks and troughs to begin with.

First of all, thanks for sharing the files for the Harman 2018 and 2019 IE targets.

I’m not exactly sure which experiment you’re referring to. My post is only about the new target curves and the SoundCheck software. Their measurement routine uses two couplers so they can measure both channels at the same time. The software isn’t trying to correct anything via EQ. They just predict a preference score.

I've got a few doubts about the Harman target itself and I had some recent communications with Sean Olive about this. Firstly, I should concede this could be an irrelevant rant from somebody who doesn't sit squarely in the middle of the bell curve. The Harman target could still be a correct average. But one thing that's always bothered me is the assumption that it's a fixed curve, even for that "average" individual, because of this:

https://en.wikipedia.org/wiki/Equal-loudness_contour

I asked Sean about this, and here was his reply:

"Yes, it is probably true that at lower levels people might prefer more bass to compensate for reduced sensitivity to bass. Our listening tests were done at normal/comfortable levels (average 80 dB (B-weighted) so it should be more representative of typical playback levels than low levels."

There's no other physical reason for that increase in the low frequencies in the Harman target curve (it's not missing as a result of the ear canals being occluded). Sean didn't mention treble, but I think that's an enigma too, because the same effect ought to apply to the upper frequencies at low volumes, and there's no physical reason that I can see for wanting a roll-off beyond 10 kHz:

As I understand it, the Harman research makes no claim what the response ought to be above 10 kHz because it’s very difficult for the measurements to get perfect agreement there from different couplers, let alone real human ears. What the 2019 IE target shows there is a guess, and it happens to look like it’s rolling off.

...unless, maybe(?) you're trying to mimic sound over large distances (maybe large hall or open-air concert) where you have the usual 1/r^2 decay of all acoustic waves but also faster decay of higher-frequency waves due to viscosity:

https://en.wikipedia.org/wiki/Stokes's_law_of_sound_attenuation

But that would seem to depend on what source you're trying to reproduce, e.g., the large concert venue vs. small studio acoustics. I'd rather that were taken care of in the recording, i.e., place the mics wherever you want and then do your best to faithfully reproduce that. Personally, I'd also like to take a little more off the lower treble of that Harman target curve (it looks like they at least started in that direction in 2019):

If anybody wants the data files for these curves, you can grab them from these links (you can then import them as an FR into REW):

https://www.dropbox.com/s/bnleuzf25m575mu/Harman_target_2018.dat?dl=0
https://www.dropbox.com/s/jr6qjz5v45swlib/Harman_target_2019.dat?dl=0

The 2018 target has a lot of fine details around 10k, which seems dubious to me. If you could plot the high and low points of a confidence interval around there, all that detail in the target curve is going to be swallowed up. What is it? +/- 10dB above 10 kHz, if not more?

The 2019 IE target loses the bump around 8 kHz. If I recall correctly, Harman used a normal IEM and EQ’ed it to test various targets. Their listeners could adjust the bass and the treble to their preferred level. I wonder if the research did anything to address the test IEM’s peak from the ear canal resonance. Could that hump around 8 kHz be the result of peaks being shifted in frequency and then averaged later? In any case, the 2019 target doesn’t have it anymore. I agree with the change—if anything, it’s more “honest” about the uncertainty. Besides, they said that listeners preferred a smooth response there, instead of tall peaks resulting from resonances.

Also, notice that they’ve lowered the peak at 3 kHz yet again. I suspect that the average and/or median preference is actually higher than this. But what will happen if an IEM has it at +11 dB like they had before? People like me who hear that resonance stronger will hear a dip when listening to a headphone tuned to the target. The people on the other side of the average will hear a peak. It’s these people who will whine louder and say that the target is too shouty. It’s because a peak is considered more offensive and less tolerable than a peak at the same frequency.

I hope this trend doesn’t go too far. A headphone that doesn’t have enough rise at 3 kHz will sound dull and uninvolving.

The latest victim of this trend? Sure, it sounds safe and inoffensive. But those of us on the other side of the bell curve might think it’s a little muffled. But just a little. (BTW, the results aren’t all that repeatable. I’m getting different frequency response measurements when I come back and measure it again after setting them aside for a while. Apparently the AirPods Pro have some kind of adaptive EQ.)​

Anyway, thanks for sharing the file for the 2019 IE target. I actually have something that comes close to it:

BTW @yuriv, as always, no pressure here, but did you make any progress with measuring our little set of headphones in that simple vinyl-tube coupler?

I posted my vinyl tube measurements on the shared folder. I used a Dayton Audio iMM-6. I’m getting a roll-off with the Apple dongle and REW for Windows 10. The roll-off isn’t there in Audiotools in iPadOS when it’s using the same mic and dongle. I wonder if Audiotools is applying its own calibration. In any case, we can fix the REW measurements later with a calibration file.

The tour IEMs reached Europe over two weeks ago.

 

[Slater]

I happened to be reading @Brooko’s review of the LZ A4, and saw a photo of LZs factory showing their FR measurement setup:



Does anyone recognize the couplers they are using? They look really big to me, like they are designed for measuring dynamic headphone drivers not IEMs. Also the sealing material looks like some sort of alginate or ballistic gel perhaps?

Just curious, that’s all. I find it interesting to get a rare glimpse of the exact setup that an individual manufacturer is using, since all we usually get to see from manufacturers are the FR graphs not the actual rig.

 

[earfonia]

Many thanks to @csglinux for the invitation to join this tour! Yes I would like to join this measurement tour. Thanks again!
I'm in Singapore (Asia) so just inform me when the ER2SE happen to land here in Singapore.
I just bought generic 711 coupler complete set with (according to the seller) calibrated microphone from AliExpress during the recent 11-11 sale. The 711 coupler will probably arrive around end of this month.

Just a bit of background, I've been experimenting IEM measurement since 2015 using MiniDSP UMIK-1 and Superlux ECM888B microphone, with DIY heat shrink tube coupler. I tried different volume and length, and now have been using 7mm diameter - 14mm length coupler as shown in the picture. Once I got the 711 coupler I will make some comparisons and probably build more heat shrink coupler that could match the 711 coupler (hopefully).

Result of measurement from different coupler length from 0mm up to 35mm in 2.5mm step:

 

[Slater]

Many thanks to @csglinux for the invitation to join this tour! Yes I would like to join this measurement tour. Thanks again!
I'm in Singapore (Asia) so just inform me when the ER2SE happen to land here in Singapore.
I just bought generic 711 coupler complete set with (according to the seller) calibrated microphone from AliExpress during the recent 11-11 sale. The 711 coupler will probably arrive around end of this month.

Just a bit of background, I've been experimenting IEM measurement since 2015 using MiniDSP UMIK-1 and Superlux ECM888B microphone, with DIY heat shrink tube coupler. I tried different volume and length, and now have been using 7mm diameter - 14mm length coupler as shown in the picture. Once I got the 711 coupler I will make some comparisons and probably build more heat shrink coupler that could match the 711 coupler (hopefully).



Result of measurement from different coupler length from 0mm up to 35mm in 2.5mm step:

Are those couplers made out of silicone eartips with heat shrink tubing over top? It’s not totally clear from the photo, and I’d love more details!