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

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

  1. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I loaded the graphs in REW and checked the difference between the two graphs for each IEM. Here's the difference between the averge difference for the 4 IEMs:
    Difference difference.jpg
    Actually damn close for those IEMs below 5kHz. 2db max. In this case you win.
    Still, there's a reason the IEC coupler mimics the volume of a real ear canal, not some arbitrary volume that gives better looking FR graphs for some IEMs. If we had 100 IEMs, not 4, we would most likely see why.

    I was also curious how the "4cc" plots would look like compensated for the average 4cc effect. Attached
     

    Attached Files:

    Last edited: Apr 17, 2018
  2. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,959
    Trophy Points:
    113
    Location:
    Irvine CA
    Those results may mean that your system is very sensitive to IEM positioning. That is bad. But that does not apply to every DIY rig.

    Furthermore, it is very misleading to claim that the IEC coupler mimics the volume of a real ear canal. That is bullshit. Every individuals ear canal is different. So in that sense, the IEC coupler is very f'ing arbitrary.

    I think you need to fix your coupler to be less sensitive to positioning. You may also have to play with volume and diminish resonances as much as possible. The IEC couplers in general still suck ballz in this regard. Which is depressing considering the "thousands of hours" spend on developing them and how much they cost.
     
  3. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Those are La Cenric's graphs from the first post. I simply took the difference between his IEC 711 and DIY 4cc results for each IEM, averaged them and then plotted the deviation from average for each IEM.

    Yes, every ear canal is different. Such is life. It seems our two ear canals are particularly different :p. Still, an average is not arbitrary.
     
    Last edited: Apr 17, 2018
  4. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,959
    Trophy Points:
    113
    Location:
    Irvine CA
    Well then that means @La Cenric's rigs (likely both the IEC and DIY) may be position sensitive. IEMs may also be volume sensitive.

    An average is arbitrary because the variance is large enough to render it arbitrary.

    Ideally one would like to see positional insensitive rigs with maybe 3 different volumes (large, medium and small ear canal).
     
  5. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    That's precisely my point.

    The ear canal isn't position insensitive though. Which is why folks with small or long ear canals won't be able to hear a TWFK without the big 8-10kHz half wavelength resonance, while people that can get deep insertion won't get the peak.
     
  6. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,959
    Trophy Points:
    113
    Location:
    Irvine CA
    If things are indeed not position sensitive, then what remains is volume and shape.

    I don't know how much the shape of the canal affect response. I'm also not comfortable claiming that position is not as much of a problem, but it is possible.

    I do know volume affects response. Again, I would consider giving at least 3 compensated responses for large, medium and small ear canal responses when dealing with IEMs. This is likely not necessary with headphones where the ear canal volume is negligible relative to the full headphone driven volume.

    EDIT: BTW, possible I have overlooked things, but I have yet to see at least 3 different "standard" large, medium and small volume IEM couplers from IEC or ITU or whatever.
     
    Last edited: Apr 17, 2018
  7. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Volume, shape and the acoustic impedances of the microphone and the measured in-ear, which both change with damping. It is position sensitive though.

    I guess to really see which coupler is most accurate we should make precise EQ profiles by ear like AstralStorm used to do and then compare those against the measurements.

    If anything the graph I posted shows that you won't really need that. It's close enough anyway.
     
  8. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,959
    Trophy Points:
    113
    Location:
    Irvine CA
    Oh crap.

    Ok, are IEM measurements in your experience position sensitive or not? Two post above I read you felt they were not.

    As far as which coupler is most accurate, f**k it. IEMs measure so badly on standard rigs that just a quick listen should suffice to at least get one in the ball park. Once things get there maybe some fine tuning is in order. But in the current state of things, measurements in many cases are grossly off. Completely useless off.

    EDIT: I mean I know what 10 or so dB of DT990 treble and/or high mids hotness is. Many IEMs that seem to measure that way or much worse do not sound at all like that.
     
    Last edited: Apr 17, 2018
  9. Crinacle

    Crinacle Friend

    Staff Member Pyrate IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    812
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    Let me see if I can clear the air a bit.

    RE: IEC vs DIY, the differences are attributed to the simulation of the inner ear, not the canal. Simulating the inner ear results in more SPL from ~700Hz up. If I were to match the resonance point of my IMM6 DIY rig to my IEC setup with the proper coupler length, I can get something that would be uncannily close when compensated since all the peaks and dips are in the right places. Example below:

    [​IMG]

    If I remember correctly, the coupler used on the IMM6 DIY rig had about a 18mm separation from IEM tip to microphone face.

    RE: Which sounds closer to subjective impressions, I would say the IEC ones sound closer to my ears, with the exception of the resonant peaks. In most cases, ignoring the resonant peaks in my mind makes the curves representative of what I hear. Now, you could theoretically compensate the 0.4cc coupler to IEC performance, but then you get weird issues like the one in the Carbo Tenore curve where the treble peak's placement and magnitude are off.

    RE: Prediction of peak location and magnitude, it's actually fairly predictable once you get a feel of things. Let's take an example of a measurement with an insertion depth that creates a resonant peak at 10k. Inserting the IEM deeper into the coupler will result in the resonant peak shifting higher up and the magnitude of said peak decreasing as well. Inserting the IEM shallower results in the peak shifting lower down and the magnitude increasing (in most cases). There is at least a consistency in the trend of peak position, though the topic of magnitude is still something I'm looking into.

    RE: What is the point of using the IEC microphone if it "looks" less accurate than the 0.4cc coupler ones? The point here being that the 0.4cc coupler is accurate for higher frequency measurement since the resonant frequency for that coupler lies beyond 20k, so it doesn't interfere with the data. However, the data below that is not as accurate as the IEC measurements since the peak placements (particularly >5k) are different and arguably not at the right places, along with the lack of inner-ear simulation of course. Which is why I am playing around with the idea of using a hybrid measurement where I show both the IEC and the 0.4cc measurements (though your main focus should be on the IEC measurement), with the 0.4cc measurement only confirming if certain peaks are actually there or not.

    RE: Are resonant peaks actually there on subjective listening? They very well could be. After all, when simulating a specific ear canal length, we're kind of trying to use the "average" canal so there are definitely some people out there who would hear the resonant peaks as shown on the graphs. The main point of this is not whether or not a certain peak is really there, per se, but rather saying "hey, that peak over there, you may or may not hear it depending on the anatomy of your canal". For my ears, I do not hear the resonant peaks as shown on the IEC measurements, but that doesn't mean that I speak for everyone else on earth. There very well might be people who hear the TF10 as V-shaped as opposed to having laid back treble, for instance.
     
    Last edited: Apr 17, 2018
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,624
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I don't hear those resonant peaks either, at least I categorize those types of peaks as false peaks.

    The frequency of those false resonant peaks are predictable on a known coupler once an initial measurement has been made with another known coupler.

    However, the frequency and magnitude, or even the existence of a false resonant peak cannot be predicted, at least without knowing the internals, of any random IEM, for any specific coupler.

    For example, I could never had predicted the frequency and magnitude of false peaks of the UERM, ER4S, Andromeda for any specific coupler without first taking a measurement on a known coupler too.
     
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,624
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Two possible explanations for false peaks on IEC type couplers:
    1. The IEC couplers and how they model the ear, eustician tubes, and mouth are wrong, insufficient, etc. As @ultrabike alluded to, real life behaviors may not be linear, and cannot be presented as tidy circuits with RLC components.
    2. The human brain is able to distinguish or adjust for ear canal resonances (specific to the human hearing the sound) because it can process in the time domain. Frequency domain measurements are essentially derived through an average of signals throughout a time interval.
    Conclusion: @La Cenric's short coupler is a winner for a single process IEM measurement solution. Content below 5kHz has 2db variance from an established IEC coupler. This difference can easily be attributable to human factors. Past 5kHz, the short coupler does not produce nonsensical false peaks which experienced and trained ears do not hear.

    I hope to build upon the ideas here. I hope others will too. I hope the objectivist IEM dweebs read this and understand.
     
    Last edited: Apr 17, 2018
  12. Crinacle

    Crinacle Friend

    Staff Member Pyrate IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    812
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    To the industry's credit, GRAS did build a higher res version of the IEC60318-4 (model RA0402) that specifically kills the resonant peaks without touching the others. It's essentially a coupler that is minimally affected by insert depth. Unfortunately it's new, expensive and I can barely justify dropping 2k on audiophile equipment, much less $20,000 on measuring equipment.

    I'll be able to gather quite a bit of data in July. I'll need a larger sample size of both IEC and small-volume coupler data before I can make any concrete claims on all this. Currently just putting this idea out there for spitballing.
     
  13. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,959
    Trophy Points:
    113
    Location:
    Irvine CA
    I really liked your post in part because you were able to reproduce IEC results, which IMO is very important to show one has a good handle on things. I have not done that.

    RE canal or inner ear stuff, it is very important to me that you clarified that the response is attributed to inner ear modeling. I don't think the inner ear is bypassed by any IEM, just as the entire ear is not bypassed by loudspeaker. As such, IMO it's effects should be removed from the measurement entirely. This may be done by calibrating it out, or by using a rig that does not imitate it's effects just like loudspeakers are not measured with an entire ear model attached to a mic.

    RE subjectives, resonant peaks cover almost the entire midrange and all of the treble. If we argue that all is good, except for the resonant peaks, we can almost throw away everything except the bass and the lower midrange. Which almost leaves us with a bunch of nothing. You are however right in that if we were to use a ruler and kind of average things up, one could get a better sense with the RED plots than with the GRAY plots. Resonant peaks not withstanding, which are pretty severe (in all cases).

    RE prediction, the location of peaks being a function of insertion depth makes sense to me. It is possible that the peaks are interactions with the chamber, and the location of possible standing waves will depend on depth. Shallower insertion may increase standing wave wavelength and decrease the frequency of the resonant peaks onset. If there was a way to completely remove standing waves through absorption w/o loosing acoustic impedance, that would be IMO an achievement.

    RE IEC vs 0.4cc, I would argue that none of them are accurate past 1 or 2 kHz. Again, I feel inner ear should be removed completely, and as early as possible. However, outer ear effects may need to be added IMO, which might be hard because outer ears are also not universal. Volume effects should be included. Given canal depth and width differences, I do not expect acoustic impedance to be universal by any measure. As an example, I tend to hear bassy IEMs neutral given I have large and deep ear canals relative to my peers.

    RE peaks correlating to subjective experience. I do not think I find them correlating to what I hear at all. However, effects of bypassing the outer ear may color things out a bit. The brain may quickly adapt to the changes though.

    Thank you for bringing all these points. What I present above is my most current and honest understanding. It may be flawed and I welcome any corrections or light into details I may be missing.

    Brain will IMO correct for inner ear. I've always been more concerned about the bypassing of the outer ear (and indeed entire head) for IEMs. Headphones obviously bypass the head. Loudspeakers bypass nothing, and so anechoic chamber measurements with omni mic may yield satisfactory measurement results (not withstanding room interactions).

    I also like @La Cenric's short coupler better.

    Removing bullshit peaks is a noble and laudable idea. However, $20k for it is not a noble and laudable idea. Cheering "standard" couplers as superior to certain DIY rigs w/o appropriate consideration of measurement results is an even less noble and laudable idea (none of which I think you are doing).

    I have a feeling that increasing the length of the coupler can yield potentially better high frequency results at the expense of bass. So maybe a short and long coupler mixture is in order. I think this is what you are pursuing, and it makes sense to me. But I would experiment with an even longer depth than what is available in the IEC deal. Just a thought. LOL! maybe it's the other way around. One would have to play with it I guess.
     
    Last edited: Apr 17, 2018
  14. Crinacle

    Crinacle Friend

    Staff Member Pyrate IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    812
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    I guess my justification for continuing to use an inner-ear simulator is simply to keep my data in line with what others' already have. Innerfidelity, Speakerphone and Rinchoi are great resources using the entire IEC suite that I like to double-check against, just in case there might be any anomalies that my measurements may present. So far, nothing odd has popped up that can't be explained by my own variations in measurement procedure.

    I think you've misunderstood me; I was referring to the resonant peak of the measurement system, not the transducer itself. The measurement rig resonant peak usually manifests itself as one singular high Q spike in the 7-10k range that varies based on insert depth and/or coupler volume (though much more heavily influenced by the former). Ignoring this peak when interpreting IEC measurements brings them much more in line with my own subjective impressions, IMO. Again, my ears do not perceive the resonance that the IEC rig has but certain ears might. They are definitely outliers of the population, but there is a good probability that they exist.

    I've experimented with long couplers before and it mostly results in a nasty, ever present 5-6k peak in every single measurement. Not very useful data, even simply as a supplement to the small-volume coupler stuff.
     
  15. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I think there's one thing we disagree about that's causing a lot of confusion:
    I don't think the peaks are entirely false or simply not there and I think it would be wrong to make a coupler that pushes these half wavelength resonances past the audioband, because the resonances are there. Now that doesn't mean that I hear such a peak in every IEM that I've listened to, but I do think it's fundamentally there, just as resonances between the earcup and ear are there with (almost) every over ear headphone. You can hear these resonances listening to tone sweeps.
    If the resonances weren't there, then how could insertion depth change treble so much? The Brainwavz B2 is unlistenable to me with shallow insertion due to a peak at 8kHz, but with special deep insertion techniques I can push it past 10-12kHz and it's fine.
    I do agree that the magnitude of such resonances is wrong with the 711 coupler.

    The reason why I think we sometimes don't hear such a peak at all is because we're used to hearing certain resonances. There's a quarter wavelength resonance from the open ear canal at around 8.5-9kHz (example) and the pinna introduces some peaks at 7-9kHz that I have seen in my own measurements depending on the angle of incidence. This can be averaged out (DF), but it's important for sound localization (soundstage in headphones/IEMs). When listening to speakers you will also hear such a peak, especially in a very damped room (link).

    If the natural open ear canal peaks overlap with the artificial closed ear canal half wavelength peaks then it's subjectively not as annoying. Sometimes the frequency is about right, but the magnitude is far too high as is the case with TWFK IEMs such as the DBA02, so we need to use deep insertion to push it to a region where it's not as annoying. Sometimes the frequency and magnitude are about right as with the RE-272 where I hear a minor peak when listening to sweeps, but with music I don't notice it.

    As I've said before I don't think the effects are entirely linear, so we cannot simply remove it without some penalty in accuracy. Sometimes this is a major effect (outer ear with headphones) and sometimes it's fairly minor like in the case of these 4 IEMs. Still, as @La Cenric said I don't think the compensated .4cc measurements correlate very well with what I hear, either.
     
  16. james444

    james444 Mad IEM modding wizard level 99

    Pyrate Flathead IEMW
    Joined:
    Jun 3, 2016
    Likes Received:
    2,100
    Trophy Points:
    93
    Location:
    Vienna, Austria
    Personally, I'm a bit skeptical towards their high res simulator, which uses a filter unit to control the 13.5k coupler resonance.

    https://www.superbestaudiofriends.o...-3-shut-the-f-up-jude.5674/page-7#post-187218

    Not sure whether that really "specifically kills the resonant peaks without touching the others". How would a filter like that be able to distinguish between resonant peaks and non-resonant peaks? Just as an example, compared to my subjective impression, measurements from this new coupler seem to show the 10k region of the IE800S as more benign as I hear it. And others have noticed similar things for the MDR-Z1R measurements... so I'm really not sure if these discrepancies are all due to individual HRTF variance, or if their filter unit might possibly tend to gloss over peaks in the 10k region in general.
     
  17. Crinacle

    Crinacle Friend

    Staff Member Pyrate IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    812
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    Test run.

    TF10 test.png
    Note that this is the TF10's left channel, that's why it looks different from the measurement in the first post

    A lot of pseudoscience going on in the form of trial and error so I'll need to fool around with different IEMs. TF10 above matched to the peak at 12k since it's the common ground between IEC and 0.4cc. Yes? No? I don't know. The combo graph does sound more like what I'm hearing, though.

    Full overlay below for transparency:
    concept overlay.png
     
  18. Crinacle

    Crinacle Friend

    Staff Member Pyrate IEMW MZR
    Joined:
    Dec 12, 2016
    Likes Received:
    812
    Trophy Points:
    93
    Location:
    SG
    Home Page:
    Yeah the RA0402 is basically the centre of the whole Jude vs Everyone Else war from back then. Not saying that it's perfect, but it least acknowledges that part of the 711's shortcomings.

    Then again, I'm looking at Jude's MD+ measurement and it seems that there's a rather large resonant peak still. Maybe it isn't as effective as I thought it would be.
     
  19. james444

    james444 Mad IEM modding wizard level 99

    Pyrate Flathead IEMW
    Joined:
    Jun 3, 2016
    Likes Received:
    2,100
    Trophy Points:
    93
    Location:
    Vienna, Austria
    There's variance in ear canal length and volume, ear canal shape, ear drum impedance and the stapedius reflex. All of these affect frequency response... and I'm not even talking about age related hearing loss.

    I have no idea whether the combined variance is large enough to render an averaged standard arbitrary. But for those who want to get a better idea of what we're up against (or those with simply too much time on their hands ;)), here's the most extensive study on these aspects that I've come across:

    https://pdfs.semanticscholar.org/77de/b9c44b683417477f62873a9b82bde27e378a.pdf

    [​IMG]

    [​IMG]
     
  20. briskly

    briskly Friend

    Pyrate
    Joined:
    Oct 3, 2015
    Likes Received:
    59
    Trophy Points:
    18
    Location:
    New York
    Scattered thoughts:
    • The reference data for IEC 711 was drawn from the two old farts willing to stick tubes in their ears.
    • Middle ear resonance frequency of 711 is probably a bit off, a bit lower in real life, but the general behavior is there.
    • 711 is intended to be a simulator from the second bend, the bony portion of the ear canal, and beyond. Geometry there is simple, a straight and slightly taped tube connected to the eardrum.
    • Even without any leak, the ear canal would lose 1 or 2dB of bass compared to IEC 711.
    • The new GRAS coupler aims to damp its own resonance a bit, but it still has position dependence, A new resonance is introduced near 20kHz to increase upper end response.
     

Share This Page