A portable rig for measurements on the go

Discussion in 'IEM Measurements - La Cenric's IEM Playhouse' started by La Cenric, Dec 25, 2016.

  1. La Cenric

    La Cenric Friend

    Staff Member Friend
    Joined:
    Dec 12, 2016
    Likes Received:
    274
    Trophy Points:
    63
    Location:
    Melbourne
    Recently I've been playing with this Dayton IMM-06, a tiny $25 calibrated mic that works with mobile devices. Currently on my iPhone 7, I'm using the app called "FFT Plot" by Onyx, though other members use more powerful apps like AudioTools and Spectrum Analyser (by Onyx) for more in-depth factors that I'm not yet interested in. Definitely not a replacement for some of the bigger, more advanced desktop rigs out there, but it does really nice quick-and-dirty measurements.

    How it looks like with the ER4PT
    [​IMG]

    Very simple vinyl tubing coupler, just enough to fit most IEM tips. I've managed to do some measurements at some audio shops here and there and it has been rather consistent with measurements from Tyll and Marvey, though not without its caveats. A few comparisons below:

    Oriolus MK2: Mine and Marv's
    MEE Pinnacle P1: Mine, Marv's, Speakerphone's
    Etymotic ER4XR: Mine and Marv's

    My observations:
    Seems like the IMM-06 with vinyl tubing is rather accurate from 20Hz to 2000Hz, though it gets a little iffy from there. Especially in the range from 5K to 10K, where the SPL it measures seems to be consistently lower than expected. Perhaps some compensation in that area would be better.

    At any case, this setup serves as a great way to do some pseudo-objective comparisons on the go. May not be the most accurate, but can definitely produce an FR curve that shows an IEM's general tonality.

    The rest of my measurements (currently compiling a written list for easier navigation)
     
    Last edited: Jun 28, 2017
    Untilted, ultrabike and Marvey like this.

  2. Serious

    Serious Immature child

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    962
    Trophy Points:
    93
    Location:
    Munich
    This looks pretty interesting. Looks like you already measured a bunch of IEMs. Do you think those curves match your subjective impression?
    It's even more portable than my WM-61A rig. Question is what kind of SNR and distortion you can get with this microphone. I've found that even two WM-61As can have different amounts of distortion.
    Part of it could be the microphone, but I think most of it is the tube diameter and insertion depth.

    BTW: Your Oriolus MK2 link links to the Pinnacle P1.
     
  3. La Cenric

    La Cenric Friend

    Staff Member Friend
    Joined:
    Dec 12, 2016
    Likes Received:
    274
    Trophy Points:
    63
    Location:
    Melbourne
    Fixed.

    The curves sometimes do match my own subjective impressions (I'm not perfect after all), though some IEMs (like the Rhapsodio Galaxy) I enjoyed more than what I would've expected.

    The FFT Plot app that I'm currently using can only output FR curves. There are more powerful apps out there that can measure THD, like Audiotools. Currently, I'm more interested only in FR curves so I decided to go the cheaper route for now.
     
  4. shotgunshane

    shotgunshane Moderator

    Staff Member Friend
    Joined:
    Sep 26, 2015
    Likes Received:
    1,140
    Trophy Points:
    93
    Location:
    The ATL
    How much vinyl tubing did you use? I'm assuming an amount equal to 2cc? Where can you pick up the tuning? I wouldn't mind playing around with this.
     
  5. La Cenric

    La Cenric Friend

    Staff Member Friend
    Joined:
    Dec 12, 2016
    Likes Received:
    274
    Trophy Points:
    63
    Location:
    Melbourne
    I have a few lengths of tubing that I got from my neighbourhood hardware store. I just ensure that there's around a 12mm space from eartip to mic face with every measurement. Perhaps some changes for deep inserts or shallow insert IEMs like the ER4 and T8iE respectively.

    The tubing I use has an inner diameter of 8mm, which snugly fits into the mic nozzle without need for acoustic putty or blu tack.
     
  6. Marvey

    Marvey Loves sex and records

    Staff Member Friend
    Joined:
    Sep 24, 2015
    Likes Received:
    26,633
    Trophy Points:
    113
    Location:
    Calabasas, CA
    Good stuff. I'm getting a general idea of a reference target curve given the relative differences between our measurements. I know that mine expect a bit of bump between 2-4kHz. Your measurements don't seem to need such a target.
     
  7. La Cenric

    La Cenric Friend

    Staff Member Friend
    Joined:
    Dec 12, 2016
    Likes Received:
    274
    Trophy Points:
    63
    Location:
    Melbourne
    A bump is actually required on mine as well. In fact with the proper compensation files and fiddling with the tubing length, the lowly IMM6 + tubing gets scarily close to 711 rigs.

    mic differences.png

    I've actually had a few chats and discussions with Luke (creator of the Veritas) and we've sort of come to the conclusion that the IMM6+tubing rig measures about the same if not identically to the Veritas (going purely off raw numbers). I'm currently developing a proper compensation file for my IMM6 based off this data and I can send it over to you for your Veritas if you'd like.
     
    bartzky likes this.
  8. bartzky

    bartzky Rando

    Joined:
    Dec 22, 2016
    Likes Received:
    13
    Trophy Points:
    3
    Location:
    Germany
    Home Page:
    I've tested two individual Veritas myself and have seen many uncalibrated results taken by different Veritas and they all vary a lot. In fact I've tried to use calibration files created for other Veritas couplers and they all were totally useless because the deviations between those individual couplers are way to high. Even though a calibrated Veritas can archive very good results, I'd say your uncalibrated IMM6 is much closer to IEC-results than most uncalibrated Veritas couplers.
     
  9. macro

    macro Rando

    Joined:
    Oct 3, 2017
    Likes Received:
    4
    Trophy Points:
    3
    Location:
    California
    Thanks for sharing, La Cenric! You have one of the most extensive sets of IEM and especially CIEM measurements I've found anywhere. Much appreciated! I'm very tempted to get one of those Dayton mics to play around with.

    I was curious to see how your modest little mobile measurement rig compares to Tyll's fancy measurement head, so I decided it would be fun to write a sloppy little program to compare them. My program scanned 10 measurement plots for earphones that you and Tyll have both measured (ER4PT, ER4XR, GR07 Bass, GR07 Classic, SD2, SE535, UM1, W20, W40, W60). Then it averaged the left and right channel of Tyll's measurements and subtracted that from your measurements to create 10 "difference curves". Finally, it averaged those 10 difference curves together and spat out this:

    [​IMG]

    Note that the vertical position of this curve on the graph is arbitrary. The measurements were level matched at around 450 Hz for comparison.

    It looks like your measurements are about 6 dB warmer below 2 kHz compared to Tyll's uncompensated measurements. There was quite a bit of variation above 5 kHz or so, though, likely due to a combination of inconsistent insertion depth and variation between individual drivers. This seems to be very consistent with your IEC711 comparison as well.

    And just for fun, here are a few CIEMs adjusted to see roughly how they might look on Tyll's measurement rig:

    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]

    If you do come up with a better compensation curve, I might be able to leverage the code I've already written and incorporate your compensation curve to make a simple command line program that scans in your FFT Plot screenshots and automatically outputs a compensated curve like the ones above. Feel free to let me know if you're interested.
     
    Last edited: Oct 4, 2017
    Marvey and La Cenric like this.
  10. La Cenric

    La Cenric Friend

    Staff Member Friend
    Joined:
    Dec 12, 2016
    Likes Received:
    274
    Trophy Points:
    63
    Location:
    Melbourne
    Amazing work. If you're willing to work on in further, I'll honestly want to use your tool and basically replace my entire database with re-compensated curves. I'm was actually working with another programmer with the exact same concept, but it's more a "in his free time" kind of gig so progress has more or less stagnated now.

    I have a compensation curve derived from direct comparison between my 318 coupler and the IMM6 rig but it never really seems "right" when applied directly onto the iOS application. Your compensation plot definitely seems more accurate at first glance. Since my main goal is to aim for an approximation rather than a dead-accurate representation (just to facilitate easier interpretation from those used to 318/711 curves) I would suggest a constant value from 3000Hz up (in your case, being 6dB all the way to 20k) in order to minimise the doubts relating to insert depth and rig differences etc..

    Do follow-up with this project. I'm very interested in seeing it come to full fruition.
     
    Last edited: Oct 4, 2017
    Marvey likes this.
  11. TomHP

    TomHP Almost "Made"

    Joined:
    Nov 24, 2016
    Likes Received:
    118
    Trophy Points:
    43
    Location:
    Taiwan
    That compensation curve seems to make sense. A 711 is more than just a tube, it has an intricate network of volumes and tubes to mimic eardrum impedance.

    The result (simulated) can be seen in this PDF on page 14 --> http://www.aes.org/technical/documentDownloads.cfm?docID=177

    Above 1.5 kHz the impedance starts to rise, resulting in a rising response, which aligns quite well with the calculated compensation curve. In fact, I wouldn't think you would be too far off if you just applied the compensation seen in that PDF to your measurements, as long as the tube you have in your coupler is similar to a 711 coupler.
     
  12. macro

    macro Rando

    Joined:
    Oct 3, 2017
    Likes Received:
    4
    Trophy Points:
    3
    Location:
    California
    Thanks!

    The program definitely needs some work before I share it with anybody. It's not user-friendly to run, and it makes some fragile assumptions about the FFT Plot screenshots. If you used a phone with a different screen resolution or if the FFT Plot developers changed the layout or color scheme of the app, it would probably break the algorithm I use to scan in the screenshots. There are a few other kinks that need ironing out as well. I will definitely try to clean it up for you when I have some more free time, though.
     
  13. Biodegraded

    Biodegraded Acquaintance

    Joined:
    May 28, 2017
    Likes Received:
    90
    Trophy Points:
    18
    Location:
    Calgary AB, Canada
    @La Cenric , does the app you're using save the plots only as screenshots? I ask because some similar ones store the underlying data in eg xml files; if these exist and could be exported, it might make @macro 's project easier.
     
  14. briskly

    briskly Friend

    Friend
    Joined:
    Oct 3, 2015
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    New York
    @macro 's curve seems in the ballpark. I guess that's a win.

    I don't think you should be lumping the dynamics and BAs together with the same compensation curves. Dynamics are often less dependent on the load impedance of the coupler than balanced armatures. BAs rely heavily on the small canal for midrange and treble response.
     
  15. TomHP

    TomHP Almost "Made"

    Joined:
    Nov 24, 2016
    Likes Received:
    118
    Trophy Points:
    43
    Location:
    Taiwan
    Not sure I follow/agree with that. How would the load impedance result in anything different for dynamic drivers than BAs?
     
  16. briskly

    briskly Friend

    Friend
    Joined:
    Oct 3, 2015
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    New York
    The dynamic would usually be more compliant, whereas the balanced armature stiffness tend be fairly high. The stiffness of a BA is enough that its displacement changes little when the earphone is placed into the ear canal, so that the volume displacement is about the same even if sound pressure is not. A simple way to test the impedance of a device is to drive it with a source with very high impedance. Measuring the resultant pressure should trace out the impedance, analogous to procedure for electric current and voltage.
     
    TomHP likes this.
  17. TomHP

    TomHP Almost "Made"

    Joined:
    Nov 24, 2016
    Likes Received:
    118
    Trophy Points:
    43
    Location:
    Taiwan
    I follow your first point about the stiffness versus volume displacement, but not sure how that relates to the ear canal resonances. A closed tube will have resonances regardless the pressure source. Isn't your point about the stiff BA contradictory to your other statement of it relying on the tube for resonances?
     
  18. briskly

    briskly Friend

    Friend
    Joined:
    Oct 3, 2015
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    New York
    Standing waves presents a pressure peak for a small amount of volume displaced at the location. In other words, the acoustic impedance of the coupler jumps up at those sharp resonances.
     
  19. macro

    macro Rando

    Joined:
    Oct 3, 2017
    Likes Received:
    4
    Trophy Points:
    3
    Location:
    California
    I like where this discussion is going. We should try to reach a consensus on what compensation curve to use as I continue making progress on the code.

    Here are all of the difference curves my program has computed for your consideration. Note that I've added the ATH-IM02 and W10 to the mix since my prior post. Please let me know if there are any other common IEMs between Tyll and La Cenric's measurements that I've missed.

    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]

    The two dynamic driver IEM (the GR07s) difference plots do appear to have a slightly cooler tilt than average, but not by a huge amount in my opinion. If anything, the ER4s look more like outliers to me, since they both have peaks around 10 kHz and dips between 10 and 16 kHz contrary to the others. My guess is this has something to do with the triple flange tips or the insertion depth of the measurements.

    Here's what the latest average difference curve looks like both with and without the dynamic drivers:

    [​IMG]
    [​IMG]

    I'm personally in favor of just using the overall average, since it's easier to have a single compensation curve that applies to all of La Cenric's measurements rather than separate curves for dynamic drivers and BAs. There really isn't a large enough sample size to generate a decent compensation curve for the dynamic drivers anyway. Not to mention hybrids. I would also prefer not to arbitrarily flatten the treble portion of the curve, but perhaps we could apply some sort of smoothing filter.

    What do you guys think?
     
    Last edited: Oct 18, 2017
    bartzky likes this.
  20. briskly

    briskly Friend

    Friend
    Joined:
    Oct 3, 2015
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    New York
    Just innerfidelity:

    DD: MDR-7550, Spider Realvoice, 1More Triple, Sphear, Adagio III,
    BA: SE846, AF160
    Others: iSine 20, XBA-Z5

    The T20 and IE800 measurements are strange. I don't know what the reference configuration of the FLC8 is either.

    I don't like that idea.
    The insertion depth necessarily relates to the treble character, given that a standing wave is induced along the length from the earpiece to the fleshy canal wall. There are fewer effective degrees of freedom, which makes alignment of the peaks doable. Sharply peaks don't sound like broadband bumps either and necessarily "ring" more as well.
     

Share This Page