Target curves and common points of comparison

Discussion in 'Measurement Techniques Discussion' started by MF_Kitten, Jul 24, 2018.

  1. EagleWings

    EagleWings Friend

    Pyrate Contributor
    Joined:
    Jul 20, 2017
    Likes Received:
    1,708
    Trophy Points:
    93
    Location:
    India
    I really like the 2 graphs overlaid on each other format. Here is my though on how response difference graphs can be a little misleading. Let's consider 2 hypothetical headphones; Bert and Ernie. Bert has a crazy 6kHz peak relative to its surrounding frequency range, but is pretty shelved down in the treble beyond the 6kHz point. Ernie is mostly linear in treble with a small bump at 6kHz.

    Ernie's response difference curve would neutralize at 6kHz and show a small dip there, but will appear brighter beyond that point. But in reality, it is possible that Bert is the brighter/harsher HP, due to the pronounced 6kHz peak. I have literally experienced this with a couple of IEMs (IEM measurements from the same person, using the same equipment).

    Bert and Ernie.jpg

    (My original post was a failed attempt in showing what I was trying to express. Thanks to @Boops for catching it. Original post can still be viewed in the quote in Boops post below.)
     
    Last edited: Jul 25, 2018
  2. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,175
    Trophy Points:
    93
    Location:
    New York
    This visualization in your third chart doesn't look quite right to me. Looking at 6k, your third chart should be showing a 20 dB difference (+10dB for Bert - -10dB for Ernie = 20dB difference), not the 0dB value you've got. Looks like you may have just added them together rather than charting the magnitude of the difference.
     
  3. EagleWings

    EagleWings Friend

    Pyrate Contributor
    Joined:
    Jul 20, 2017
    Likes Received:
    1,708
    Trophy Points:
    93
    Location:
    India
    Oops! You are right. Thanks for pointing it out. I edited my original post now.
     
    Last edited: Jul 25, 2018
  4. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    Maybe having some "difference" graphs based on a common reference (650's and 800's), and the most likely competitors (DT880 and HD600, DT770 and M50X, HE500/560 and LCD2, etc) would be good ideas?

    My main concern is that some cans might look dark but sound neutral, some may look neutral and sound thin and bright. Tiny bumps can sound huge, some treble spikes are massive but inoffensive while some are cute and tiny but shred your ears. I was not prepared for how harsh and bright the Fostex T50RP's were. They looked like they would be okay, but they sound like tin foil unmodded.

    I like the idea of having some way of showing what things sound like compared to what we all know.
     
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,623
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I nominate LCD2C and HD600 as two of the references.
    For accuracy on flat plate couplers, it's a matter of applying a B&K target with a 0-5db depression from 1-4kHz depending upon the headphone. Some headphones take advantage of pinna and concha gain more than others. The MiniDSP EARS results (with compensation so that perceived neutral is a straight line) I've found to be more accurate and have been and will be providing these going forward. Just not too many interesting headphones, so not much data based on the EARS.

    The GRAS gear can be compensated similarly (perceptive neutral as flat), but the current compensations that Jude uses actually looks more similar (has a target curve more similar) to the plate coupler results here.

    The flat plate coupler is more forgiving of placement, more consistent with seal, and does not have any high Q resonances so that CSDs can be produced easier. Flush mounting the microphone to avoid concha and ear canal seems to eliminate the high Q resonances.
     
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,623
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I would like both. Overlaid and difference plots (to common well-known headphones as reference). Really doesn't take up much more real estate.

    Where I would like input is which headphones to use as reference?

    My preference is for headphones that are somewhat normal sounding. Seems pointless to have a V shaped headphone as a reference as a V shaped difference plot will be notable when comparing to neutralish headphones.

    And no HD800 either. Although itI is well known, its screwiness (and we should consider stock only, not modified) disqualifies it.

    I'd rather use Mr. Speakers Ether standard or HE1000 than the HD800 as a reference.
     
    Last edited: Jul 25, 2018
  7. E_Schaaf

    E_Schaaf MOT: E.T.A Headphones

    Pyrate BWC
    Joined:
    Mar 28, 2017
    Likes Received:
    9,589
    Trophy Points:
    113
    Location:
    Staten Island, NY
    Home Page:
    Too bad LFF paradox / code-x aren't more widely heard, given they would make for excellent references methinks.

    It seems like most people, regardless of tonal preference, seem to think either of these two are within a standard deviation of normal-sounding at any given point in the FR. Never heard a subjective impression of either that indicates otherwise.

    LCD-2C and 650 are both widely heard though (so might make good references), though I wouldn't really call LCD-2C
    normal-sounding personally.

    In terms of FR, Clear / Eikon / Auteur might be good options as well. At least better options than stock HD800.
     
  8. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    7,952
    Trophy Points:
    113
    Location:
    Vancouver BC
    Like this? In the top graph, red is HD6XX; in the bottom one, HD6XX is a straight line at 85 dB.

    [​IMG]

    [​IMG]

    Procedure in REW: export the reference curve (HD6XX in this case) as txt, then in Excel eliminate everything but the freq and SPL columns, and subtract the latter from SPL at your reference frequency (in this case, 85 dB @ 1 kHz). Then save it as tab-delimited text, and use it as a mic calibration file on the headphones you want to compare.

    I thought it'd be a simple matter of using REW's 'trace arithmetic' in the controls menu, but I was getting flaky and inconsistent results every time I tried...

    Edit: slight screwup in the second graph - I did the arithmetic on an already mic-calibrated version. This is better, with HD6XX as the red line for reference. Note that applying per-octave smoothing has the effect of introducing some error at higher frequencies...

    [​IMG]

    Edit 2: Ok, final final now... Calibration curve made from unsmoothed HD6XX FR (red) then all smoothed at 1/12:

    [​IMG]
     
    Last edited: Jul 25, 2018
  9. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,175
    Trophy Points:
    93
    Location:
    New York
    Playing a bit more with the visualization using this example.

    sbaf_measurement-mock_lcd2c-he400i@2x.png

    Edit: tweaks for legibility
     
    Last edited: Jul 26, 2018
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,623
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'm stealing all of your ideas for the next gen measurement visualizations.
     
  11. Elnrik

    Elnrik Super Friendly

    Pyrate
    Joined:
    Jan 10, 2017
    Likes Received:
    8,973
    Trophy Points:
    113
    Location:
    Denver CO
    Home Page:
    Zmf Auteur / Eikon / atticus?

    I think the reference headphones should not just be mid-fi / popular on Amazon / commonly found everywhere type headphones. There should be a few diamonds in the rough types included. Headphones that are genuinely great regardless of category or price range, regardless of how unique or one-off they are.

    I really don't care how a new headphone compares to the Beyerdynamic line (990,880,770, etc...), as they don't exhibit any genuinely unique qualities other than murderous treble. Now, go find me another headphone that sounds like the Atticus, Eikon or Auteur. Or the Zenith PMX2, or LFF Code Sex, or HD 650 KISS.

    You get the point.

    Sure, we need easily obtainable, dime a dozen, and well-known headphones to serve as a reference point or else the comparisons lose their meaning. I understand that. But we should definitely throw some unique character headphones into the mix.
     
    Last edited: Jul 25, 2018
  12. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,175
    Trophy Points:
    93
    Location:
    New York
    Fine by me.
     
  13. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    I agree on LCD2 and 600/650 as the reference points, at least for planars and open dynamics specifically. Thry are the gold standard benchmarks for those categories. I personally nominate the DT770/80ohm (not 250 or 600) as the closed back reference for the same reason, as I see those pop up in conversations about good closed cans very often, and many people own them. You could use the M50X here too, but I don't often hear as much positivity about them from people with some experience.

    Going for the most common cans in each category seems to make sense to me, since it'll be the most recognized references.

    It's very very hard to find a perfect "TOTL" reference, since there are so many weird-ass measurements from that sphere. They all have very eccentric ideas of what "perfect sound" is, and where the focus should be, who the demographic is, etc.

    My main request is simply to go for what most people will have heard, what most people own and know, in each category. What do people want to compare to. I think the ones I listed are gunna be the most common gold standards in each category, with the exception of planars maybe being more dominated by Hifiman owners because of the price. Fostex planars don't count because they're the exception and not the rule in that field.
     
  14. Hrodulf

    Hrodulf Prohibited from acting as an MOT until year 2050

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    4,248
    Trophy Points:
    113
    What about LCD-2 variance? HD600/650 is fine by me as thousands of people have them and they have almost no variance sample to sample.
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,623
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    LCD2C variance is OK from three samples. Maybe take the average of them.
     
  16. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    I think it would be fine to smoothen the reference graphs so the tonal balance is the focus and not the individual squiggles.
     
  17. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,623
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Difference plots should be 1/3 octave smoothed.
     
  18. MF_Kitten

    MF_Kitten Banned per own request

    Banned
    Joined:
    Jan 1, 2016
    Likes Received:
    181
    Trophy Points:
    43
    Agreed completely
     
  19. k1arg

    k1arg Facebook Friend

    Contributor
    Joined:
    Oct 12, 2015
    Likes Received:
    153
    Trophy Points:
    43
    Location:
    London, UK
    Re sample variance, it would be really interesting to have some rough idea how many pairs Sonarworks had to measure before the distribution of measurements indicated that the 3dB accuracy claim had been hit. Say for LCD2s and HD650 for comparison. :p

    Also about sonarworks, an older version of the software supported simulating an HD650. Looking at the compensation curve for various phones in that mode seems to give a comparable plot. And similarly factors out their target curve.
     

Share This Page