1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Etymotic ER2SE - A Reference Headphone for your Ears and Your Couplers

1 2 3 4 5
7 8
  1. castleofargh Contributor
    https://arxiv.org/ftp/arxiv/papers/1811/1811.03389.pdf that one comes with a big decision to try and position the rig at a fixed distance(to get a specific resonance from all subjects). so added to all that cool work, we have to consider the extra variable that is insertion depth and total ear canal length. so basically the situation is worst than what they show.

    I had another one that was quite complementary to that study, but the link is dead(that's what I get for bookmarking instead of downloading everything). from not so reliable memory, it was saying something very intuitive. like how beyond the length and total volume, the variations in shape didn't impact the low freqs much but started to have more impact above 1 or 2kHz(I think) and rapidly becoming quite significant as we increase the frequency.
    Slater likes this.
  2. SilverEars
    Anybody have the mdat file of the latest Harman ie target? Also, diffuse field target curve as well?
  3. yuriv
    If we’re already measuring the frequency response at 94 dB SPL at 500 Hz with a sweep in REW, then we’re also already making basic distortion measurements at that level. Look at the difference:

    Apple ME186LL-A harmonic distortion, 94 dB SPL at 500 Hz measured by sweep.png
    Apple ME186LL/A harmonic distortion measured with sine sweep, 94 dB SPL at 500 Hz​

    Apple ME186LL-A harmonic distortion, 94 dB SPL at 500 Hz measured with stepped sine.png
    Apple ME186LL/A harmonic distortion measured with sine stepped sine, 94 dB SPL at 500 Hz, 12 points per octave, 2 averages

    The the distortion at 3 kHz for this model is well known. (See the measurements at InnerFidelity, ClarityFidelity, and Rin Choi’s blog, for examples.) But if we measure at 80 dB SPL, we don’t see this behavior.

    Apple ME186LL-A harmonic distortion, 80 dB SPL at 500 Hz measured by sweep.png
    Apple ME186LL/A harmonic distortion measured with sine sweep, 80 dB SPL at 500 Hz.

    We’re not driving the IEM hard enough for it to show the nonlinearity. I think we might find the same for the vast majority of IEMs. Moreover, we lose resolution and precision in our measurement because the background noise is relatively higher. Fortunately for those listening to this IEM, pure, sustained 3 kHz signals at 94 dB seldom occur in real music. I think it sounds just fine.

    94 dB SPL is a high average level for music listening. I listen at a much lower average level, depending on the dynamic range of the track. If it’s classical music, the average will be in the mid-low 70s. But the peak levels can go well above 100 dB SPL if it’s playing to match the level of a live performance.

    Maybe we’re basing our judgment on test tones at 1 kHz. Those are annoyingly piercing at 94 dB SPL. It’s even worse at 3k. But try listening to a 50 Hz tone. I find it moderate, as long as I don’t listen to it too long. This doesn’t surprise me, given what we know about equal-loudness contours and our sensitivity to hearing low frequencies. The bass in pop music easily exceeds 94 dB from time to time. So do the drums in a full orchestra. We want our headphones to reproduce bass at that level without any annoying distortion.

    Anyway, the tour has some inexpensive IEMs for distortion testing. I’m using a source that’s limited to 1 Vrms. In addition, REW’s default level for sweep testing is -12 dB, so I’m not too worried about damaging our inexpensive test subject with this particular measurement rig.

    IEC 60268-7 does specify an alternative test level: 1 mW at the headphone’s rated impedance. Spec sheets for most IEMs show a higher sensitivity than 94 dB SPL at 1 mW. Off the top of my head, I can remember only the ER4S and ER4B being less sensitive. For most IEMs, 1 mW is a more severe test level.

    Put another way, 94 dB SPL is sending much less than 1 mW to the IEM. That kind of power isn’t going to melt the voice coil or the surrounding material. So should we be concerned, instead, about over excursion, which is when the driver is pushed past its mechanical limit? IEMs have the pressure chamber effect, so the diaphragm’s excursion should be constant until you get to a couple of kHz, where it start decreasing. If 94 dB SPL bass doesn’t sound unreasonably loud and distorted, then the small voltages shouldn’t be enough to make the driver bottom out at any frequency. Otherwise it'll definitely show up in the distortion measurements, and you'll hear it.

    That said, it hasn’t prevented one of us from destroying an IEM with a powerful amp designed to drive inefficient cans. Nothing is ever guaranteed, but that’s what the tour’s beater IEM is for.

    A possible drawback of testing at 94 dB SPL with some of these couplers is that the performance of the measurement mic can become a factor. When testing IEMs that can play back loud with really low distortion, it may come to a point where the measurement shows the mic’s distortion instead of the IEM’s.

    ER4PT harmonic distortion, 80 dB SPL at 500 Hz.png
    Etymotic ER4PT harmonic distortion measured with sine sweep, 80 dB SPL at 500 Hz

    80 dB SPL. Nothing to see here. Let's see what happens 14 dB higher:

    ER4PT harmonic distortion, 94 dB SPL at 500 Hz measured by sweep.png
    Etymotic ER4PT harmonic distortion measured with sine sweep, 80 dB SPL at 500 Hz

    The 3rd harmonic (orange) dominates until around 1.2 kHz. This matches other measurements published online. What doesn’t match is the rise in the 2nd harmonic distortion (red) from 1 kHz and up. Is the IEM distorting or is it the measurement system? BTW, this measurement was taken using 1M points, 8 sweeps. It produces smoother traces.

    Here are two more examples:

    LG Quadbeat 3 AKG harmonic distortion, 94 dB SPL at 500 Hz measured by sweep.png
    LG Quadbeat 3 AKG, harmonic distortion measured with sine sweep, 94 dB SPL at 500 Hz

    Sony MH755 harmonic distortion, 94 dB SPL at 500 Hz measured by sweep.png
    Sony MH755 harmonic distortion measured with sine sweep, 94 dB SPL at 500 Hz

    Measurements published online show much lower 2nd harmonic distortion and THD for these two models. So is this really showing the distortion of the measurement system? I got similar results with the Startech and the Apple USB-C dongle, so maybe it’s the mic. I want to see how the tour’s throw-away IEM measures on other rigs. If one of them measures really low THD, then I’ll know for sure. In fact, I say make an additional measurement and push it to 100 dB SPL like Tyll did (and what Brent Butterworth at Soundstage Solo does).

    Even if the measurement shows mostly the measurement system’s distortion, it’s still useful information. On my measurements at 94 dB, the combined THD didn't exceed 0.5%. It means that the IEM’s THD can’t be worse. The measurement also shows that it’s mostly the second harmonic, which is the least offensive kind.

    The variability at 1 kHz is probably part of the reason the IEC has the standard test level at 500 Hz. The tolerance is less than +/- 1 dB at 1 kHz, so one of my couplers is probably out of spec there. The IEC coupler’s first internal volume causes the ~5 dB rise in the response near 1 kHz. This is why those Dayton Audio iMM-6 vinyl tube couplers don’t show the 3k peak as high in their measurements: no internal volume with slits, no Helmholtz resonance, and therefore no rise in the response. After about 2 kHz or so, the IEC couplers are in rough agreement again—until about 8k or so. IEC 60318-4 doesn’t really dictate what the response ought to be much above that, but they do give the typical curve. Two couplers could be 4-5 dB apart at 10 kHz and both be in spec. The cheap clones that most of us have probably do the minimum to stay within tolerances— and then roll off above 8 kHz.

    Jaakko Pasanen posted them here: https://github.com/jaakkopasanen/AutoEq/tree/master/compensation

    In REW follow the menus: File/Import/Import frequency response, then select the .csv file.
    Last edited: Jul 25, 2019
    csglinux and hakuzen like this.
  4. castleofargh Contributor
    if you're really motivated to get some answers and references for your setup, the shortest and most reliable path IMO would be to purchase a few Knowles or Sonion BA drivers, and see if you can replicate their specs(Sonion usually comes with pretty THD graphs). they use 0.1V, so here is yet another reference standard to consider. :weary:
    my circumstances won't let me do anything relevant for THD, between a lot of noises, how my rig is way too DIY to come close to IEC anything, and a mic that's not made for that purpose. plus my super crappy soldering iron associated with my inferior dexterity would probably require that I purchase 10 drivers to end up with usable contacts on one after wiring(it hurts because it's true). but as the industry leaders for those types of drivers, I don't think it would hurt to rely on their specs.
  5. SilverEars
    Anybody know of a good adapter solution for measuring earbuds? I guess one of the problem is keeping the earbuds in place on the coupler with adapter. An adapter that simulates outer ear where buds keeps in place.

    This is why some couplers are mounted on a base vertically with a hinged cover that swings down on top of the earphone to keep in place. Because iems in the coupler can move depending on the tips.
    Last edited: Jul 25, 2019
  6. hakuzen
    replace the in-ear adapter of your coupler with the adapter for earbuds (if you got it).
    use putty/blue-tac to keep the earbud in place.
    i made a DIY press, similar to that hinged cover which comes with some coupler stands, but it added excessive distortion, even after adding sorbothane feet to the press. guess the only way to avoid that disto is having the press attached to the stand (impossible for me). so the putty is better solution.
  7. SilverEars
    I don't have the adaptor? Where do you get it, and how does it look like?

    Also, have you measured a custom before? Trying to figure out how close to the mic the custom outlet should be. @csglinux suggested taking the adaptor off.
    Last edited: Jul 25, 2019
  8. hakuzen
    some sellers provide them as an option when purchasing the coupler. i don't remember if you can purchase the adapter alone; i can't search for them now, you can surf in taobao to locate it (search for iec60318-4 or iec711), although taobao is updating their platform and browsing isn't much pleasant now.

    here are the earbuds adapter, and the balanced armatures adapter (tubing needed).
    this shop adds the adapters by default, search into their shop to check if these accessories are sold alone: https://item.taobao.com/item.htm?id=38609814212
    [​IMG] [​IMG]

    this shop provided the accessories separated in same product page, but it seems they are out of stock of them or don't do it anymore. pity, because they were very cheap. search into their shop as well. https://item.taobao.com/item.htm?id=539623117975
    in this pic, you can see a bevel where the earbud stays:
    [​IMG] [​IMG] [​IMG]

    edit: i forgot it, you add the adapter together with in-ear adapter, like you can see in these pics..
    Last edited: Jul 25, 2019
    csglinux likes this.
  9. hakuzen
    csglinux likes this.
  10. SilverEars
    Now that I've imported the Harman ie compensation curve onto REW to create mdat file, more question. Lol. Hopefully all you readers are getting helping hands from the experts on this thread as well.

    Anybody know how to compensate the iem measurement with the Harman or diffuse field targets?
    Last edited: Jul 27, 2019
  11. antdroid
    Use the EQ Wizard in REW. Use the House Curve feature in settings and open your Harman Target curve as a text file. I'll attach, then play with the wizard. :)

    https://redirect.viglink.com/?forma...ssion - Head-Fi.org&txt=Targets-Reference.zip
  12. csglinux
    Lots of good, thought-provoking stuff here. I can certainly take another shot at re-measuring THD at 94 dB before posting our little package on to Jude - and we can then recommend others to take THD measurements at 80 dB and 94 dB? A couple of concerns here:
    1) I might be encouraging others to destroy their headphones. Or the tour headphones. Please don't destroy headphones :)
    2) There could be a problem with a low-distortion headphone (e.g., ER2SE) of not being able to distinguish distortion effects from the coupler mic, pre-amp & sound-card and distortion effects from the headphone itself. As you suggest @yuriv, this is probably still fine for comparative purposes (since distortion components would be additive), but it might make it tricky to reach an overall conclusion on THD levels of a very low-distortion headphone. I've found the dynamic mics in most of these 711-clone couplers to be not all that good at measuring THD.
  13. hakuzen
    time line for upgrading my mobile game has expired (with no success, btw..). so i can start to prepare the mini pc now. guess i'll use JDS C5D as amplifier; if i get enough time to properly calibrate it, will try THX 789 amplifier as well (even lower noise and disto).

    summary of planned measurements:
    94dB@500Hz, 24bit 96kHz, 3 or 4 sweeps of 1M, or 2 sweeps of 2M, self-averaged (this helps to decrease noise, good for distortion measurements), FR and THD (will also try 80dB for disto to check differences), sensitivity and impedance curve, 3 different insertion depths (reference plane, deeper, and shallower insertion, fixed with putty), common tips rolling, possibly 100dB@500Hz (although higher loudness perception is not reflected in FR). won't do various output impedance simulation, because the test iems are all dynamic drivers. results in REW .mdat format + pics.

    please tell me if i forget something. for example, response to square signals, if available in REW (i haven't tried it before); frequencies?
  14. csglinux
    Quick status update: I received the set of tour IEMs back from @SilverEars. (Thanks buddy!) They'll be on their way to @jude within the next few days. I believe @yuriv is next after that.

    Our little IEM collection seems to be increasing rapidly - I confess I've not gotten around to measuring all of them. I did revisit the ER2SE and MH755 in order to re-measure THD at 94 dB (all my reference points now @ 500 Hz, following the suggestion from @yuriv) and I've also measured @SilverEars' AKG IEMs. I've learned a few interesting things over the last couple of weeks that made me revisit (and re-do) some of the measurements I put on the first post of this thread. I think some of these points are worth sharing.

    The background here was me trying to compare against a set of reference data I had for my ER4XR. The ER2SE unfortunately don't come with reference measurements from Etymotic, but the ER4XR do. It had puzzled me for a while that I couldn't get very good agreement here. Specifically, Etymotic measured a peak at ~ 13 kHz, whereas I measured a trough:


    I contacted Dave Friesema at Etymotic (Dave used to go by the handle @EtyDave when he was on headfi) to ask about this and he gave me a couple of useful pieces of information. First, as @castleofargh mentioned, Etymotic don't actually measure using eartips, but rather an "eartip simulator". This is because an eartip simulator allows them to quickly and easily get repeatable measurements, i.e., there's no risk of variance in insertion depth. Here's Dave's description of their eartip simulator:

    I begged to be allowed to buy one of these, but (not entirely surprisingly) the answer was no :frowning2: However, Dave did give me some useful information. He told me that the eartip simulator was designed to simulate the large, clear tri-flange tips, seen here on the right (note that I'd been measuring with the small tri-flange tips on the left):


    BTW, the ER2SE do NOT come with these large, clear tri-flange tips; the ER2SE only come with two sizes - small and large - of the frosted tri-flange tips, shown left and center of the above figure. (The large clear tri-flange are even larger than the large frost tri-flange that come with the ER2SE.) My initial attempts at measuring my ER4XR with the large clear tri-flange tips were better, but still a bit off. I then discovered that I could get fairly close if I didn't install the eartip all the way down the stem of the nozzle. If I left a fraction of a mm gap, things improved. Here's me adjusting both insertion depth and how far down the stem the eartip sits, i.e., the size of installation gap left between the IEM body and the base of the eartip:

    Still not 100% perfect, but it now looks like I'm measuring the same IEM. Of course, all this has got to be a practical issue from listener to listener - i.e., there are many reasons why individuals might hear the same headphone (even the exact same unit with the exact same eartips) differently in the treble. I noticed that when I received the ER2SE back, the two sets of eartips were installed like this:

    As Cosmograf once said, "Please mind the gap between life and death" :wink: As with the ER4XR, any gap also creates notable differences in the frequency response in the treble for the ER2SE. Here's a rough spread of what I can get by adjusting the eartip (S and L frost tri-flange, large clear tri-flange and tip->nozzle installation gap):

    Basically, the treble is all over the map, even in cases where the primary resonance peaks line up at the same frequency. Unless you want to go absolutely bananas and measure every possible combination in the known universe, I recommend only using the S (small-sized) frost tri-flange with the ER2SE, install the tip all the way down the nozzle stem and shoot for a ~10.5 kHz primary resonance peak. Note that that's NOT what I originally recommended - I originally suggested shooting for a primary resonance peak at ~11.5 kHz for the ER2SE. (Thankfully, everybody who's measured the ER2SE so far was smart enough to completely ignore my suggestion!) Here's why I now think that was a bad idea....

    I found that by trying to achieve a resonance peak at ~11.5 khz, the ER2SE's tri-flange eartip ended up beyond the reference plane of the coupler, and there was a risk it could get pushed up against the grill in front of the coupler mic. (No surprise that if that happens, the FR will change.)

    All this has been a long-winded way of saying watch out and be careful, because variations in eartip type, installation on the IEM nozzle, and/or insertion depth can create a crapshoot situation if you're trying for consistent measurements in the treble.

    On the subject of trying to get consistent, repeatable measurements in the treble - I also have some comparisons now from GRAS' hi-res coupler to share (on the ER2SE, @McMadface's MH755 and @SilverEars' AKG headphones). More on that in a later post...
    yuriv, McMadface, Raketen and 3 others like this.
  15. csglinux
    Below are some comparisons against the GRAS 'hi-res' coupler. GRAS were kind enough to loan me a demo RA0402 (pre-polarized mic version). They asked for feedback on it, so perhaps we can send our collective thoughts and impressions as I'm obviously a pretty limited sample size on my own. I realize that getting feedback from the whole group might require a second round of these tour headphones, but what might be interesting to know is: which of the measurements below do you think gets closer to what you actually hear when running a sine sweep with those particular headphone sets? (I have an Android app called "Frequency Sound Generator" on my LG V40, which has both coarse- and fine-grained control over sweeps of various waveforms, including sine waves. I'm not sure, but I presume there must be an equivalent for iOS?)

    Tour unit ER2SE:

    @McMadface's MH755:

    @SilverEars' AKG:


    Note that because of the differences in simulated canal resonance amplitudes, THD measures mostly lower with the hi-res coupler:
    er2se_thd_hi_res.png mh755_thd_hi_res.png akg_thd_hi_res.png

    Externally, it's identical to a standard 711 coupler:

    20190811_150218~2.jpg 20190811_150252~2.jpg

    I'm going to hold off on my own thoughts and comments for now as I'm still talking to GRAS about this (and desperately trying to understand the implications of different input and transfer impedances) and it might be good to first get some confirmation (or otherwise?!) from @jude on these hi-res coupler measurements. (N.B. Even if the standard 711 coupler measurements get closer to what we hear, there could still be benefits for designers in minimising the influence of canal-length-related resonances on an output graph.) Anyway, feel free to pm me if you'd prefer your thoughts/comments to be passed on anonymously.

    P.S. So far, only tour participants have access to the database of all measurements, but at some stage we might want to compare results from all contributors. What do our contributors think of making a master plot for each headphone, just to show the spread from the various measurements? (We could leave the plot legends blank or ambiguous?)
    Last edited: Aug 14, 2019
1 2 3 4 5
7 8

Share This Page