MiniDSP EARS: Deriving SBAF compensations from MiniDSP files?

Discussion in 'Measurement Techniques Discussion' started by purr1n, Nov 7, 2018.

  1. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    I don't remember who asked me for this, but I am finally posting this information. I am hoping that folks with EARs interested in using the SBAF headphone compensation will be able to derive it from the HEQ files provided by MiniDSP.

    https://docs.google.com/spreadsheets/d/1uUqkpD3EJMzpl8FeJbNOkjApvMKAIL3IP8Y1At_ldng/edit?usp=sharing

    Version 1.1 (with screw removal): https://www.superbestaudiofriends.o...nsations-from-minidsp-files.7067/#post-259898

    See related threads:
    https://www.superbestaudiofriends.org/index.php?threads/calibrating-minidsp-ears-part-2.6880/
    https://www.superbestaudiofriends.org/index.php?threads/minidsp-ears-measurement-rig.5377/

    Here are the deltas of the custom SBAF FR compensations for my EARS to the MiniDSP provided compensations. They are by no means perfect; but I feel comfortable enough running with them now. A few things to keep in mind:
    1. I have no idea if adding back the SPL difference will actually work with your individual EARS. The L and R differentials derived from files provided by MiniDSP were quite a bit different from my L and R and differentials. I am very concerned about this.
    2. One issue I had was to get every headphone to look right, to reflect what my ears are telling me, but this seems impossible. Tweak one thing that makes one headphone's measurement look right, and it screws up another headphone's measurement. These compensations are a compromise and balancing act. At the end of the day, it's good enough and it works given the limitations of the EARS which is a budget product.
    3. Frequency response plots derived from the SBAF EARS compensations will be very different from other plots. The target for perceptual neutral was intended to be a flat horizontal line straight across the graph. It's very important to keep this in mind.
    4. The plots here were derived from multiple measurements of over a half dozen headphones. Because of inconsistencies in the positioning process, I applied weighted averaging of results. Some headphones are more immune to positioning variances. Other less so. Data which I felt was obviously bad was thrown out or re-acquired.
    [​IMG]
     
    Last edited: Jun 1, 2019
    WHO23, Ardacer, eIndi and 10 others like this.
  2. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    Here was my first test case AFTER the compensations were developed. This headphone I did not have on hand to develop the compensation, so it a good test case. My concern still lie with L/R channel matching.

    ZMF Verite - thick pads.
    [​IMG]
     
    WHO23, ultrabike, zach915m and 2 others like this.
  3. Vtory

    Vtory Illogical Spock

    Friend MZR
    Joined:
    Sep 12, 2016
    Likes Received:
    3,794
    Dislikes Received:
    36
    Trophy Points:
    113
    Location:
    TX
    Great! Did you not correct phase info? Or, just use what minidsp corrected for phases?
     
  4. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    Just use miniDSP provided phase.
     
  5. Vtory

    Vtory Illogical Spock

    Friend MZR
    Joined:
    Sep 12, 2016
    Likes Received:
    3,794
    Dislikes Received:
    36
    Trophy Points:
    113
    Location:
    TX
    Compensation comparison (sbaf vs mine v3).jpg
    (No smoothing)

    Tried to apply sbaf compensation curves to my elex measurements. Because my EARS were surprisingly well-matched between left and right channels (note: I very recently purchased ears), applied marv's delta after averaging left and right.

    Honestly a bit surprised that difference is much less than expected.
     
  6. Denonic

    Denonic Rando

    Joined:
    Oct 7, 2018
    Likes Received:
    1
    Dislikes Received:
    0
    Trophy Points:
    1
    Location:
    Australia
    Having issues with channel imbalance after calibrating mine to the SBAF compensation. I do have one of the EARS from the latest batch that they just restocked. Will work on it a bit more and get screenshots when I'm back at home.

    Also does the pinna do something weird to the HD800 peaks? They seem to be in all the wrong places on the measurements I've found.
     
  7. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    HD800 or anything with angled drivers are going to be super sensitive to positioning. Can you try another headphone?

    How off are the left and right FRs of your EARS? Yeah, screenshots will help.
     
  8. Josh83

    Josh83 Friend

    Friend
    Joined:
    Mar 1, 2017
    Likes Received:
    506
    Dislikes Received:
    1
    Trophy Points:
    93
    Location:
    Charlottesville
    My EARS were also from the recent batch, and applying the SBAF compensation threw my channels out of whack, too. The difference in the original HEQ in the SBAF spreadsheet is larger than the modified HEQ in the SBAF spreadsheet. Obviously it must've worked for Marvey's EARS, but not mine, which had a smaller initial channel difference.

    The supplied HEQ for my EARS for a HD6XX looked like this:

    [​IMG]

    Applying the SBAF compensation:

    [​IMG]

    What I did from there was take the channel that looked most correct (in comparison with Marvey's measurements of the same headphone with EARS, as well as Tyll's measurements and others), apply that to both of my EARS channels, then adjust the incorrect channel (for example, if you ended up applying the left SBAF compensation to both the L and the R, you would tweak the R) based on the initial differences between the channels. Then I continued to tweak from there to hopefully continue to get it closer to symmetrical.

    Here's the result (apologies for the color switch):

    [​IMG]

    Basically, it just took a little tweaking. But the SBAF compensation is way more intuitive than the supplied one, and it's worth the effort even if the initial application throws off your channel balance.

    Note that you'll need to measure a bunch of headphones (preferably ones known for good channel QC), then try to triangulate your tweaks so that the channels look closest to in balance across the whole range of headphones.
     
    Last edited: Jun 24, 2019
  9. Denonic

    Denonic Rando

    Joined:
    Oct 7, 2018
    Likes Received:
    1
    Dislikes Received:
    0
    Trophy Points:
    1
    Location:
    Australia
    So here is the supplied HEQ calibration measuring my 58x.
    [​IMG]

    Here is the SBAF compensation:
    [​IMG]

    I'll work on channel matching the way Josh83 suggested, but from the looks of it, My graph still doesn't look close to the 58x you posted above. Maybe I'm calculating it wrong?
     
  10. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    Close enough. There are some familiar patterns. It's not unheard of that Sennheiser puts in running changes that alter the FR slightly. Also, the EARS can be slightly different from unit to unit. Because of the high Q resonances, there may be particular sensitivies where a generic calibration based from my EARS won't work exactly die other EARS.

    The best way is to establish a calibration headphone(s) and send it out.
     
    WHO23 likes this.
  11. Denonic

    Denonic Rando

    Joined:
    Oct 7, 2018
    Likes Received:
    1
    Dislikes Received:
    0
    Trophy Points:
    1
    Location:
    Australia
    [​IMG]

    Managed to match it a bit better. Just needed to play around with adding a static value difference to push the majority of it up and then adjusting where it looked a bit much. Still looking much better than stock! Will measure some others when I get some time to see how they stack up. Thanks for the taking the time to do this. For people just getting into measurements with the EARS, this is invaluable!
     
    WHO23 likes this.
  12. eIndi

    eIndi Rando

    Joined:
    Apr 8, 2019
    Likes Received:
    22
    Dislikes Received:
    0
    Trophy Points:
    3
    Location:
    Europe
    Thank you for your work on this. I have settled for a mix of HPN and SBAF compensation. HPN below 1.3 kHz, SBAF above. Here is a comparison of my HD800 with HEQ, HPN; SBAF and Hybrid:
    [​IMG]
    I went with HPN below 1.3 kHz because I think SBAF would show a bit too much bass there. My HD800 seems to be pretty good in terms of sub-bass, but I doubt it has the same level at 20 Hz as it does it 1 kHz.

    Some more comments on the compensations:
    • HEQ has an elevated bass target and that clearly shows. Good to EQ to, but not to measure and show deviatons vs neutral.
    • HEQ shows way too much treble.
    • HPN is garbage between 2 and 5 kHz.
    • HPN shows too little treble above 9 kHz.
    • SBAF treble seems about right.
    • SBAF shows a bit more upper mids/lower treble between 2 and 4 kHz vs HEQ. I think SBAF is correct here. Trying to EQ to flat using HEQ makes this frequency range sound overemphasized.
    • Not sure if what compensation is best between 4 and 9 kHz. That area is quite affected by the EARS resonance. Note how the 6 kHz HD800 peak does not show up at all.
     
    WHO23 likes this.
  13. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    One thing to keep in mind with the SBAF compensation is that it was arrived at using over a dozen headphones.

    Over the course of developing the compensation, I would tweak the response of one headphone to get it to look better, only to get the response of another headphone to look worse.

    Ultimately, it was a balancing act of give and take, to arrive at a curve that didn't look too far off for the dozen+ headphones used as the basis. In addition, there were also attempts to find and cancel out consistent repeated behaviors that could be attributable to the EARS.

    The thing I did not want not do is rely on a compensation that kept changing (Olive/Harmon, cough cough). I wanted something that would work good enough for any not yet measured headphone that I threw at it.
     
    WHO23, Lyander, eIndi and 1 other person like this.
  14. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    https://docs.google.com/spreadsheets/d/1uUqkpD3EJMzpl8FeJbNOkjApvMKAIL3IP8Y1At_ldng/edit#gid=722506266

    This is the version 1.1 tweak that requires removing the screws on the coupler. The screw removal wasn't the panacea that I expected. In most cases, getting a good seal was easier; but in a few others, then open holes seemed to make getting a good seal more difficult. It's still a similar deal. Attention needs to be paid to where the screws or in this case, the screw holes are in relation to the pads.

    Also, a few other minor tweaks were applied.
     
    WHO23 and Vtory like this.
  15. Vtory

    Vtory Illogical Spock

    Friend MZR
    Joined:
    Sep 12, 2016
    Likes Received:
    3,794
    Dislikes Received:
    36
    Trophy Points:
    113
    Location:
    TX
  16. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    49,387
    Dislikes Received:
    80
    Trophy Points:
    113
    Location:
    Antarctica
    Yeah. Those were 1.1, hence the adjustments in the mids on the HD600 during that process.

    1.1 without a screws should be largely compatible with 1.0 with screws. The minor tweak was a small adjustment just below 9kHz of a db or so.

    No screws seems to be better with some planars, more consistent, more repeatable. But worse for HD6xx series.
     
    Walderstorn, Lyander and Vtory like this.
  17. MLegend

    MLegend Friend

    Friend
    Joined:
    Oct 4, 2015
    Likes Received:
    96
    Dislikes Received:
    1
    Trophy Points:
    28
    Location:
    Northwest Florida
    When performing measurements, do you set your dac's sampling rate to 48000hz to match the microphone or does it not matter? Does it matter what type of amp and dac you use when taking measurements?
     
  18. WHO23

    WHO23 Rando

    Contributor
    Joined:
    Feb 23, 2019
    Likes Received:
    21
    Dislikes Received:
    0
    Trophy Points:
    3
    Location:
    Singapore
    I purchased the MiniDSP EARS a few months ago. I haven't used it much because I never got any graph to look remotely right with their included compensations.. until I found this thread.

    I just added the Purr1n's 1.1 difference onto my my raw compensation files without doing the averaging method. It seems to me the slopes looks pretty good for the most part. Nicely done Purr1n! However for me I have two following issues:
    1) the bass is elevated for all measurements
    2) 1kHz to 4kHz looks off, especially for diffuse field target headphones.


    MiniDPS Measurements (1.1 SBAF Compensation) vs Sennheiser Official Measurements
    [​IMG]
    [​IMG]
     
  19. WHO23

    WHO23 Rando

    Contributor
    Joined:
    Feb 23, 2019
    Likes Received:
    21
    Dislikes Received:
    0
    Trophy Points:
    3
    Location:
    Singapore
    Edit: Just to clarify, Purr1n's compensation worked amazing for most of the headphones I measured (such as the DT1990, Verum 1, LCD-2). For me it was hard to gauge the treble quantity of diffuse field tuned headphones (such as the ADX5000) so I wanted to just compare them to my HD800S (which I use most often).

    Note: My compensation is useless for anyone who don't have a HD800S. I didn't do an average of multiple measurements and of multiple headphones. Also for compensation of the elevated treble for any individual, it depends on the position and angle of the speakers you usually listen to so this is a futile experiment.
    [​IMG]


    I've taken some cues from elndi and just used raw measurements for 10-1300Hz range.After the change I was still unhappy with with the frequency response for the HD800S at the 300-4500Hz range. I then overlaid my current compensated results and changed the values so that they look similar to the provided Sennheiser HD800S diffuse-field compensated frequency response.

    HD800S frequency response (R):
    Raw 10-1300Hz & SBAF 1300-20000Hz (red) VS Official Sennheiser (blue)
    [​IMG]

    HD800S frequency response normalized at 300Hz (R):
    Raw 10-1300Hz & SBAF 1300-20000Hz (red) VS Official Sennheiser (blue)
    [​IMG]

    HD800S frequency response normalized at 300Hz (R):
    Raw 10-300Hz & 300-4500Hz tweak & SBAF 4500Hz-20000Hz (red) VS Official Sennheiser (blue)
    (black) is is the new compensation curve that's the result of these two tweaks
    [​IMG]
     
    Last edited: Jul 11, 2019
  20. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    8,890
    Dislikes Received:
    29
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    But why? That Sen DF target is useless. Absolutely no bearing on how the HD800, or headphones in general, sound.
     
    eIndi, Lyander, WHO23 and 1 other person like this.

Share This Page