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

General IEM Measurements Discussions

2 3 4 5 6 7 8 9 10 11
  1. Otto Motor

    This thread is dedicated to frequency responses (and other) measurements of earphones by any Head-Fi member, using any technique and/or software! It also serves the discussion of rig setup and equipment such as couplers, microphones etc. It further encourages the discussion and exchange of compensation spreadsheets. And it promotes discussion of the tricky correlation of measurements and sound quality. Last but not least, Head-Fiers can show the visualizations of their modding steps and results and production flaws (e.g. miswiring and channel imbalances).


    Frequency-response graphs are entirely quantitative but not qualitative. They cannot tell you whether an earphone or headphone sounds good, but they may tell you whether it sounds bad.

    This writer says it well: “Here at XXX most people believe BOTH measurements and subjective comparisons are valuable together and intrinsically intertwined. The intersection and diversion of the two are where the interesting discussion and actual learning happens.”

    Frequency-response graphs carry important information of an earphone’s DNA. They visualize sound quantity and assist with…

    • Getting a basic characterization (“flavour”) or an earphone:
      • Early warning of sibilance, harshness, hardness, and piercing
      • Risk minimization: matching of earphone and personal preferences before buying
      • Information on bass and treble extension, and midrange recession
    • Quality control (channel matching; detecting internal mis-wiring)
    • Comparison of earphone flavours (by graph superposition)
    • Guiding the different steps of modding
    The graphs per se don’t tell you anything about sound quality such as timbre, particularly the degree of sound realism (e.g. accurate reproduction of an orchestra), as well as attack/decay/transients, layering, separation, detail, dynamic range, and soundstage.

    But, in some cases, the FR graphs can give indirect evidence of quality: for example, huge upper midrange peaks at around 3 kHz can be indicative of low-quality drivers (they have to compensate for a thin lower midrange, which can make the sound overly harsh and aggressive). It may also point to sibilance (7-10 kHz peaks) and fake resolution (15 kHz area peaks).

    Nevertheless, frequency response in the context of a reviewer’s description of tonality and their known preference (or your own), should give you a good idea whether a certain earphone will likely appeal to you or not.

    Last but not least, while frequency response graphs may not give conclusive evidence about the tonally, they can frequently tell you where to stay away from.

    In summary, frequency response measurements are complimentary to one’s listening impressions and should therefore be a standard staple of every review imho.

    Last edited: Mar 29, 2019
    pstickne likes this.
  2. Otto Motor
    Last edited: May 20, 2019
    surfgeorge likes this.
  3. CoiL

    Search tool word for easy finding -> coilinfoil
    Last edited: Apr 3, 2019
    pstickne, Makahl, MrDelicious and 2 others like this.
  4. csglinux
    711-compliant coupler, raw/uncompensated:

    N.B. Double-gunmetal filters were used on this particular measurement:

    Measurement procedure is described here: https://www.head-fi.org/threads/audio-measurements-on-a-headfi-budget.893084/

    Unless otherwise stated, all measurements were made with REW and no octave-band smoothing was used in the output.
    Last edited: Jun 24, 2019
    hakuzen, CoiL, Makahl and 1 other person like this.
  5. Otto Motor
    Why do you hide your graphs?
  6. Slater
    I really wish I could get one of those 711 coupler/microphone setups. I’m in the US and I’ve tried using Taobao but I just can’t get it working right with any Google translators. I mean, sure I can read the description but I’m talking actually creating an account and buying something/checking out. :frowning2:
  7. csglinux
    Nice job @CoiL! Can I suggest everybody does what @CoiL has done and include, with your measurements, a description of the coupler and eartips used? These are critical to the FR. Also, raw, uncompensated measurements would be the easiest way of comparing measurements between rigs. That's still going to be a challenge, but that's going to be what makes this thread fun :)

    Because when you have so many measurements, many metagbytes of unwanted images are going to load in the reader's browser and they'll have a heck of a time finding the one that they want. (Go back and take a look at my list now, and you might see what I mean.) If you were feeling adventurous, it would be really great if you could maintain a master list of http links to the all individual measurements, in alphabetical order, on your first post. That way, people could quickly dive into the particular measurements they're interested in.

    You should. @Otto Motor - you should too. This is the single biggest upgrade you could make to your rigs, and these aren't expensive anymore. Links to the various couplers are given here:
    https://www.head-fi.org/threads/audio-measurements-on-a-headfi-budget.893084/ (follow the hardware spoiler link). The problem with the Veritas, MiniDSP and Dayton-mic couplers is that no one compensation curve is going to exactly match the 711 coupler results for all IEMs. You'll get close with some averaged compensation curve, but there will still be some differences between ported and non-ported IEMs. The only tricky part about ordering from TaoBao is that the items initially ship to a distribution center in China and you then need to make a second payment for international shipping. I need to say a big thanks to @crinacle for walking me through that process :)
    Slater and Otto Motor like this.
  8. CoiL
    Always wonder why ppl (some reviewers) don`t include detailed info on their graphs? For example 1/3 smoothing vs. 1/24 smoothing looks VERY different and ppl might get different impression from IEM based on graph only. I will stay with 1/3 smoothing as it shows more real info, though, it may "throw off" some ppl buying IEM as they are not so experienced reading graphs.
    Or do You think I should also add diffuse field graphs? Imo I shouldn`t bother with it as, like You said, it can be inaccurate. Though, which is the best way (correct) to apply DF compensation?
  9. castleofargh Contributor
    I usually rely on a wall of text below the graph with 20 thousand warnings about not trusting the graph in any way, you know the sort of stuff that nobody ever reads.
    about compensation, I like it raw :imp:

    talking about the EARS, last week I was trying to add layers between the mic and the silicon ear, so that the resonance would be closer to 3kHz instead of being more around 5kHz by default which makes absolutely no sense. I found the right distance, tried a bunch of materials but kept failing to get a seal, so I made something with some thermo-tube thingy to go around the mic and seal things on that end properly, but after moving the mic around too much, one wire broke free. :weary:
    so I took my soldering iron and after a few minutes, I had successfully removed the 2 other wires while failing to attach the third one, and burned the mic probably beyond recovery(and my thumb a little). we have a winner! \o/
    anytime I see people soldering BA drivers and other tiny stuff, it reminds me how much I suck at this(and how dumb I was for deciding to get lead free solder when I already suck with the easy one). so now I'm scared of trying to do the mod on the other ear. I've effectively achieved negative productivity over the week. oh well, maybe I'll be brave again next week.
    hakuzen and Slater like this.
  10. kadas152
    Hey guys I've noticed that lot of you are using shorter coupler and thus moving resonance point beyond 8k opposed to what crinacle does with his measurements normalising insertion depth to 8k resonance point. Do you have any specific reason for that?
  11. CoiL
    Typical distance from mic (with Dayton iMM-6) would be 12.7mm due to average human ear-channel size? Or am I doing it wrong?
    I wondered why crinacle has so many peaks with different IEMs right at/around 8kHz.
    But why? For example with IT01 I do not hear such peak @ 8kHz and with my measurement it`s more real according to what I hear (I`m sensitive to 6.2-7.5-8kHz). If IT01 would have such peak @ 8kHz as crins graph shows, I would hear it for sure! But I don`t, no matter what tips, even without grills.
    Last edited: Apr 1, 2019
  12. CoiL
    @Otto Motor Is this measured by BD with compensation or raw?
    Looks very good to me and sounds accordingly good as my v1 does. Still best KZ to me (have no newer ones after ZSN, which is way worse than ZS5v1 imho) :)
  13. Otto Motor
    Last edited: Apr 1, 2019
  14. csglinux
    Yes. A very specific reason.
    I've mentioned this issue before on @crinacle's thread. I prefer to just seat all IEMs at whatever depth naturally results from the given size of the tip and taper on the coupler nozzle. I don't fight to achieve an 8 kHz resonance peak; you'd never do that when listening to an IEM - you'd just insert to whatever depth is natural for your ears and then decide whether you liked them or not.

    There has to be an implicit understanding that resonance peaks are going to vary from person to person anyway as a result of differences in ear anatomy. But it's important to know 1) that resonance peaks exist and 2) roughly where they are (higher frequency peaks are typically better, because our hearing gets less sensitive up there and energy content tends to eventually roll off up there anyway).

    If we were to artificially fix all resonance peaks at 8 kHz, we'd never know how 2) varied from one IEM to another.
    CoiL likes this.
  15. CoiL
    Since that 12.7mm is rough average, will explain why most measurements taken from that distance have peaks further at the end of spectrum. What I`m wondering though is the shape of human ear-channel vs. tube (circular) shape. I have taken impressions from my own ears and my small ear-channel is rather oval-flat shaped after tip edge - wonder how this affects resonant peaks? Will making coupler "average" shape of ear-channel make noticeable differences in "average perceived" resonant peak locations? Interesting subject...
2 3 4 5 6 7 8 9 10 11

Share This Page