Trying to address high frequency peaks in IEM measurements: A halfway decent solution

Discussion in 'Measurement Techniques Discussion' started by La Cenric, Apr 17, 2018.

  1. La Cenric

    La Cenric Friend

    Staff Member Friend IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    736
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    Been trying to fiddle around with this issue and have been reading up on the existing solutions and standards. Of course, this issue has been mitigated with the release of GRAS new RA0402 but I want something that people like me can use without dropping thousands on a small (though admittedly very high quality) hunk of metal.

    The background for this of course stems from the fact that most couplers suffer from resonance (even the old IEC711/318-4 which was why GRAS made the RA0402) so certain peaks, while appearing on the measurements, does not appear on subjective listening because our human ears do not have the same resonance points. And so, we need some sort of way to see if certain peaks are because of system resonance or are inherently there based on transducer output and not the measuring rig itself.

    This presentation by Knowles provides great insight to this (starting at the 22:45 mark) so give that a watch if you got the time. Hearing aid companies already use procedures based on IEC TS 62886:2016 to measure the higher frequencies of drivers, which is a system that essentially has a resonant point that's so high up that it doesn't interfere with the important data range (in this case, being 20Hz to 20,000Hz).

    If you would take a look at my existing IEC60318-4 measurements, you'd see that there's always a familiar peak at around 8-10k that shifts with insert depth. That would be something that should be ignored, but I need something more that just "because I say so". As a supplementary measurement, I decided to try to use the IMM-6 microphone with a much shorter coupler to sort of replicate the 0.4cc coupler used in the high frequency measurements based on 62886. Here's a small test sample for the "double coupler" procedure.

    These are IEMs that I have on hand with me right now so I apologise in advance if they are unfamiliar models to some of you. Red represented measurements using the IEC60318-4 while grey represents measurements using a IMM6 and small volume coupler.

    TF10.png

    First, the TF10. An undoubtedly warm IEM with an odd treble peak at 7k on the IEC rig that I do not hear on subjective listening. Testing on a small volume couple (outlined in grey) shows that the peak does not exist.

    IT01.png

    Here is the iBasso IT01, a Chinese dynamic driver IEM. The main issue on the IEC graph is the large peak at ~8k. On the small volume coupler, there is no peak whatsoever. 8k doesn't sound emphasised to my ears either.

    TFZ King Pro.png

    Here is something that does show up on the small-volume coupler: the TFZ King Pro. Also a Chinese dynamic driver IEM. On the IEC graphs, there is a "twin peak" of sorts, a sharp ~8k peak and a smaller one at ~9k. However, on the small volume coupler measurements, only the 9k peak is shown, which suggests that a 9k emphasis would show up on subjective listening still. Do take note that this 9k peak did not show up on the TF10 or IT01 small-volume coupler measurements either so it's not likely to be a function of resonance.

    Carbo Tenore.png

    Here is another interesting IEM with a "twin peak" in the higher frequencies: the Zero Audio Carbo Tenore. However, the peak placement is odd in the small-volume coupler measurement, being at ~7.5k as opposed to the IEC's peaks at ~6.5k and ~9k respectively. I am inclined to say that the lower peak is the "real" one, but I'm not too sure.

    I'll look into this further when I got the free time. For now, do give me your thoughts.
     
    Augmentin, Serious, Dino and 8 others like this.
  2. Kunlun

    Kunlun cat-alyzes cat-aclysmic cat-erwauling - Friend

    Friend IEMW
    Joined:
    Sep 26, 2015
    Likes Received:
    5,296
    Dislikes Received:
    2
    Trophy Points:
    113
    Location:
    Meow Parlour
    Good for you for trying to address this. Misleading measurements are super problematic and I don't think having a footnote to ignore above a certain point is helpful because most people seem to take any graph as holy writ.

    As for your idea, look at the tf10 graph. You are getting a big null with one and a big spike with the other. That null is not the good thing you seem to be saying it is. Nulls are no better than peaks if they are measurement artifacts. It certainly doesn't accurately "show the peak doesn't exist". It shows that something is going on, somewhere, with something. Looks a little like the second appartatus throws dips into the graph at near where the first shows peaks.

    Still, I'm glad you are working to address this!
     
    Serious likes this.
  3. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    The truth of the matter is that your el-cheapo short length coupler > anything IEC blah blah blah or GRAS.

    The only reason people (the simplistic dumbass pseudo objective ones - the euro fascist types) will NEVER believe that your el-cheapo coupler is better is because it doesn't have the letters I E C or G R A S on it. This is a problem that I've faced in the past seven years every since CS was started.

    Do I give a shit? Absolutely not. Those too lazy to try to understand the nuances of methodology, think outside the box, and only want black-and-white "knowledge" spoon fed to them from an "authority" do not deserve it. I blame this on our educational system. Colleges and universities want to pretend they are they teaching kids critical thinking when they are not.
     
    Last edited: Apr 17, 2018
    Imraan, msommers, ultrabike and 3 others like this.
  4. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    The nulls might be able to be compensated with. There seems to be some consistency with them.
     
    ultrabike and Kunlun like this.
  5. briskly

    briskly Friend

    Friend
    Joined:
    Oct 3, 2015
    Likes Received:
    59
    Dislikes Received:
    0
    Trophy Points:
    18
    Location:
    New York
    An analogy is attempting to isolate speaker response from that of its position in a room, anechoic vs. in-room response. Almost certainly for the best, especially since the ear resonance is more damped than the classic 711 spec.

    But an IEM always will be placed in one end of a tube, and there is a more limited ability of the end user to make adjustment. You can apply filters, swap sleeves, or adjust insert depth. When Knowles does a small coupler measurement, they usually aren't dealing with a finished product.

    Second peak around 13kHz (red line) is the 711 resonance, which shouldn't exist in the first place. The lower one is the canal length, which I believe is the issue being addressed.
     
  6. La Cenric

    La Cenric Friend

    Staff Member Friend IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    736
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    The null on the TF10 actually exists. Currently on mobile so I can't directly link it to you but you can take a look at Rinchoi's and Innerfidelity's measurements, both of which are made at reference plane whereas I do mine at roughly 3mm away from reference to simulate the ear canal, kinda-sorta. The problem with that approach is that it creates a resonance point of around 8k (depending on the distance from reference plane) so the resonance sort of masks the existing null, if you get what I mean.

    I should be able to get a measurement with the null when I place the IEM at reference plane, which will shift the resonance point closer to 10k. Will try that when I have the time.

    EDIT: Unless you're referring to the "null" as in the difference between the IMM6+coupler versus the IEC mic, in that case the main difference is the IMM6 rig only simulates the canal whereas the IEC microphone simulates both the canal and the inner-ear. The IEC graphs look more linear because of the inner-ear simulation.
     
    Last edited: Apr 17, 2018
    Serious and Kunlun like this.
  7. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    13,363
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    If it doesn't say <insert famous name equipment here> then it receives the grand poo pooing from the very folks you describe. Totally agree about "educational system" not providing even rudimentary critical thinking skills. It has been standardized out of the system. There are some who manage to attain an education despite the system.
     
  8. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,871
    Dislikes Received:
    0
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Not really since the FR can be way off depending on the IEM measured and it's not something you can compensate for. Different IEMs will react differently to the much smaller volume. You can even see this in the three measurements above. I much prefer looking at IEC 711 curves than my el cheapo plastic tube coupler. RE-400 and UERM are quite similar up to 2kHz on a 711 coupler, but not on my plastic tube.

    @La Cenric I quite liked the graph from the Knowles presentation with the 0.4cc coupler overlayed from 4kHz on (with levels matched). I just wish they'd shown a measurement of their IEM on the new coupler.

    One thing I tried with my plastic tube is using damping material to dampen the resonances and then compensating for the effects of the damping material on the measurements. Not really an optimal solution. Now I just use comply tips for my mic, but that doesn't dampen the resonances enough.
    Another thing I used to do when measuring in-ears is take measurements without a tube just holding the IEM close to the mic to see if resonances are real or not. But I guess using a smaller volume works similarly. Still, the ear canal does have some insertion depth related resonances, too.
     
  9. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I'm not sure the FR cannot be compensated for. From the above plots and eyeballing, it appears that they can.

    I can think of no thing worse than the false peaks where the magnitude and frequency varies of those IEC and GRAS couplers.

    Perhaps another way may be to mate two different types of measurements.
     
    maverickronin and Kunlun like this.
  10. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    The fact that some find it acceptable that standard measurement equipment should yield such wildly frequency response graphs that do not even remotely correlate with perceived sound is incomprehensible to me.

    Furthermore, the fact that some find it acceptable to use of significantly different compensation curves (which obviously yield wildly different results) to correct for the immense shortcomings of standard measurement equipment is also incomprehensible to me.

    A standard coupler should be designed that correlates with perceived sound as much as possible assuming at large, medium and small canal volume.

    Your setup is shit because it's not the 0.4cc coupler and does not meet IEC 60318-4 or IEC-60711 supported by G.R.A.S nor IEC 318, IEC 711, or ITU-T P.57 supported by Bruel & Ajaer with the sole intention of making the world better again.
     
  11. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I will add that certain DD IEMs, presumably vented, externally or into an internal volume, will produce different results between a vented coupler vs. a sealed coupler. If we think about it, our ears are vented, with varying impedances dependent on frequency. This is where the IEC couplers win.

    This is why I couldn't get proper measurements for Sony EXxxx series and needed to rig up a vented coupler to measure the CA DD and hybrid IEMs.
     
    Kunlun likes this.
  12. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,871
    Dislikes Received:
    0
    Trophy Points:
    113
    Location:
    near Munich, Germany
    No need to eyeball since we already have this:
    http://www.superbestaudiofriends.or...-for-measurements-on-the-go.3578/#post-169562
    Some massive differences even in regions not affected by insertion depth.

    Also from eyeballing, why is it that two IEMs have less bass on one coupler, while the others don't? No compensation will fix that. I highly doubt it's seal related. At 3kHz the differences vary between 3 and 6db, at 5kHz between 6 and 10db, etc.

    The peaks are predictable, always in a certain frequency range depending on insertion depth. FR below the peaks is unaffected. I find that better than wildly differing FRs for the bass and midrange depending on IEM acoustical output impedance. In other words I find the region below 5kHz more interesting than above and it's more important to me to get that right.

    @ultrabike of course it'd be better to not have to deal with such peaks, which is why there's a new IEC coupler in the first place. I don't think I can DIY a better coupler than the new IEC one. Maybe if I spent tens of thousands of hours on it I could, but for IEMs I do like the IEC couplers and I find them to correlate nicely with what I hear, definitely better than my cheapo plastic tube. Dummy heads for headphones are a different story.

    For target curves we've had a decent standard for ages: DF. Obviously it's not perfect, but everyone will likely want a slightly different FR anyway.

    Anyway, my point was really that for IEMs IEC curves correlate better with what I hear than DIY rigs, even "compensated". And at this point both the IEC 711 and the DIY rigs have massive insertion depth related peaks, so I don't really get your point.
     
    Last edited: Apr 17, 2018
    bartzky likes this.
  13. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I disagree with you. From your link: http://www.superbestaudiofriends.or...-for-measurements-on-the-go.3578/#post-169562

    I see very similar patterns below 3-5kHz with only minor differences in low bass that would easily be attributed to seal or variance based on human factors. After 3-5kHz, a lot of those peakages is purely attributable to coupler. But the patterns are mostly the same: flat out to 800Hz, mild dip 6db down by 2kHz.

    Not sure what you are saying or if I agree. I personally have NEVER been able to predict the location and magnitude of false peaks with any IEM I have not measured yet. There are too many factors that go on inside the IEM for us to make a prediction of the peak. I can only say that peaks tend to shift upward in frequency with shorter length tubing, and that longer length tubing shifts the peaks downward. But I would not dare predict how any random IEM will react in a 2.5" tube compared to a 1.5" tube.

    I see anomalous behavior of past 5kHz peaks, especially the false peaks, as being extremely harmful to the hobby. At least whlie we have idiots like k.e. or the IEM dweeby objectivists who take them as truth.
     
    Last edited: Apr 17, 2018
    ultrabike likes this.
  14. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,480
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    You haven't heard and measured that many IEMs though.
     
  15. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    @Serious, I have been in a few standards meetings. They tend to be mostly political and a venue for companies to push their solutions. Designers seldom get sent there, unless things are slow. Usually, old and semi-retired folk go there, and they spend most of their time browsing the interwebz until they hear something that may compromise the best interests of the company they are with. Believe me, they are highly political. At least all of the ones I know and attended. Most of my colleges (and myself) dread going to them.

    As far as IEC standard equipment measurements with whatever standard compensation correlating better to what I hear than DIY rigs, I completely disagree. I don't even know how you came up with that conclusion.

    Look at the results on the first post using a standard IEC60318-4 coupler. They are shit. Looking at them I have a real hard time understanding what the system will sound like. Even w all the issues, I think the gray plots make way more sense.

    Also, fuck DF curves.
     
    Last edited: Apr 17, 2018
    Vansen likes this.
  16. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,871
    Dislikes Received:
    0
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I admit the region below 500Hz and the general pattern is very similar, but even ignoring the GR07 Classic and ATH-IM02 there are some broad 3-4db differences. Most interesting to me is the ER4XR vs ER4PT. I highly doubt it's measurement variance or product variation here.

    Yes and that's not what I meant, but generally there will be a major peak at 6-11kHz depending on insertion depth. The peaks are related to geometry just like standing waves in a room with speakers are, so you could predict them.

    Maybe I haven't paid attention, but when did that happen? The Andromeda discussion was pretty annoying, but I don't think it was based solely on the measurements. As far as Andro vs PP8 goes I really can't say.

    That's true, I only have measurements of about 10 IEMs on my rig and I haven't listened to that many more.

    I gave an example with the RE-400. On my coupler the RE-400 has 5db more 2kHz than the UERM, while on an IEC 711 they're pretty much the same from 20Hz to 2kHz. I definitely don't hear such a big peak. On my RE-272 the difference at 2kHz is more similar (but it still measures with more 2kHz than on an IEC coupler, guessing about 3-4db), but an octave higher up I get a 5db peak at 4kHz while on an IEC coupler there's no peak here relative to the UERM at all.* Other DD IEMs like the Carbo Tenore measure more similarly on my rig to an IEC 711. Random 5db peaks in the most important region of the midrange are simply unacceptable to me.

    What are the measurements you are referring to when comparing a 711 coupler vs a DIY straight tube?

    *I know using HFM IEMs isn't exactly the best example. Two RE-262s I measured had a 2db FR difference.
     
    Last edited: Apr 17, 2018
    bartzky likes this.
  17. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    WTF!!!! How else do you want me to make it fucking clear to you. I don't give a fuck about your RE-400 crap. Look at the damned measurements in the first post of this thread and tell me the results using a IEC60318-4 rig (RED results) correlate better with what you hear than the ones using a non-standard IMM6 + small volume rig (GRAY results).

    This is not the first time you kind of write as if you knew what the fuck you are talking about, and have no fucking clue. My patience is not limitless.
     
  18. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,871
    Dislikes Received:
    0
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Well, then I misunderstood. I still stand by my claim that you can't compensate for the smaller volume with a simple FR compensation.

    I understood this as the IEC 711 vs DIY rigs like Marv's rig or La Cenric's earlier measurements NOT the simulated 4cc DIY rig from this thread which is totally different.

    As far as these graphs go I'd still prefer an overlay from 4kHz on like I said in my first post rather than using the smaller coupler for the whole FR.
     
    Last edited: Apr 17, 2018
  19. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    Why?

    I'm not sure what you are talking about. Per @La Cenric's post:

    I would have to look at the IEC 711 measurements. But it is clear in this thread that IEC60318-4 coupler based ones are shit relative to the DIY rig ones. Do you agree?
     
    Last edited: Apr 17, 2018
  20. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    Sorry folks. I try not to blow up like this. But this is not the first time I have a combo with @Serious.
     

Share This Page