Calibrating MiniDSP EARS Part 2

Discussion in 'Measurement Techniques Discussion' started by purr1n, Sep 27, 2018.

  1. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    9,514
    Dislikes Received:
    19
    Trophy Points:
    113
    Location:
    Irvine CA
    Octave can be such a drag. But here are CSDs of the LCD2 before and after compensation:

    Before:

    LCD2_CSD_before_comp.jpg

    After:

    LCD2_CSD_after_comp.jpg

    To keep things honest as much as possible, same windowing function, and normalized the individual FFTs to the response closest to DC. Ranges should be the same, and tried to keep plot aspect ratios similar.

    EDIT: IR filtering compensation is proly something that most folks using a ear, mannequin, or head torso simulator are missing in order to do compensated CSDs.
     
    Last edited: Sep 28, 2018
    Serious and Lyander like this.
  2. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Compensations for MiniDSP - iteration 3.0. I like these better than the last ones. HD800 actually looks bright bow.

    newcomps v3.png
     
    Vtory, Serious, Sqveak and 4 others like this.
  3. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    More here. And HE-5 (stock) looks slightly wonky.

    newcomps v3-2.png
     
    Last edited: Oct 6, 2018
  4. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Here's an example of how the comps differ with the HD58X.

    GREY = RAW
    PURPLE=MiniDSP supplied compensation
    YELLOW=v3 compensation (custom)
    comps comparo.png
     
    Last edited: Oct 6, 2018
    Vtory, Walderstorn, Lyander and 2 others like this.
  5. Magnetostatic_Tubephile

    Magnetostatic_Tubephile Friend

    Friend
    Joined:
    Dec 27, 2017
    Likes Received:
    772
    Dislikes Received:
    44
    Trophy Points:
    93
    Location:
    GoogleChrome, EU
    That 4.5-5.0k peak is probably impossible to compensate for completely. At least visually it seems to still translate somehow to pretty much all the posted graphs.
     
    Lyander likes this.
  6. Walderstorn

    Walderstorn Friend

    Friend
    Joined:
    Mar 20, 2016
    Likes Received:
    893
    Dislikes Received:
    4
    Trophy Points:
    93
    Location:
    Europe
    Kinda surprised with the ESP95x, in a good way (i never heard them). The Ether on the other hand is what i thought (and disliked) about them.
     
  7. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Yeah. I noticed that after I did all the plots. Need to investigate this. A narrow high Q drop at 4.7kHz of a few db should do the trick.
     
  8. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,195
    Dislikes Received:
    40
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    One of my concerns with the MiniDSP EARs is the width of the coupler. As best as I can tell, the average human head breadth is about 5.82-5.9" and the bitragion breadth somewhere between 5.47-5.5" (including male and female across various nationalities).

    The flat part of the EARs coupler seems to be about 5.438" in width. Plus, the whole this is just flat, except for the ears.

    Of course, it's really tricky to compare ugly, lumpy, angular heads against perfectly symmetrical, flat couplers, though it seems the EARs at best is just shy of an average bitragion breadth and nearly a half inch shy of the full head breadth. That doesn't sound horrible, but I always found myself feeling like the EARs wasn't wide enough to adequately capture a good seal on headphones, throwing off the results.

    Yeah, and the thickness and pliability of the ears relative to real ears makes a difference to in this, especially when measuring on-ear headphones.

    All that said, I'm curious how your new results would look with a couple rubber bands used on the headphone cups to simulate greater clamp. I'm guessing it will have some positives and some negatives. I just know that, personally, some of those results look like they're not getting the right seal.

    I preferred my EARs results with some rubber bands, and I actually learned that trick while watching Tyll measure a couple headphones in person. But the trick is to make sure the rubber bands are applying a very subtle force so as to not overdo it.

    Then again, I'm biased with my own results, since I still use my own ugly head for measurements. And the breadth of my head seems to be a bit over 6". I'm naturally more likely to get a stronger seal through greater clamp than the average person, for better or worse. It sure makes fitting glasses a real PITA.
     
    Lyander likes this.
  9. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I actually adjusted the bass to plot lower than the last iteration. Which ones am I not getting a good seal? FYI, the Ether I have is slightly screwy between the two channels. One rolls off massively - the one shown here. I'd post the better channel, but wanted to work on the L one first. There is a 2-3db dip with LCD2C from 30HZ below, but this is insubstantial.

    Otherwise, the bass roll offs for stuff like the Grado, HD58X, ESP950 seem like how I hear them. Keep in mind that I never get a complete seal, there is a spot under the back of my ears, under the bone, the upper back-side neck area, which prevents a total seal.
     
    Last edited: Oct 7, 2018
  10. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,195
    Dislikes Received:
    40
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    It may just be the Ether throwing me off. Thanks for pointing out the L channel issue. Perhaps needs new pads? I heard Dan talking to someone yesterday at RMAF that they have since redone the pads for the whole Ether line with new material and foam to address bass drop and the like. I guess the original stuff limited initial performance, didn't hold up over time, and led to more variability overall.
     
  11. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Just swapped pads L and R. Yeah, it's the damn pads.
     
  12. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,195
    Dislikes Received:
    40
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    LOL, isn't it great?
     
  13. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    Fixed. Compensations attempt #4
    Comp 4.png

    The only thing I am wondering now is if the 2-4kHz region should be boosted on the plots (taking into account that perceptual neutral target is flat line across). I tend to tune and prefer my speakers with a small dip from 2-4kHz, so maybe these do work.
     
    Last edited: Oct 7, 2018
    Sqveak likes this.
  14. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,871
    Dislikes Received:
    0
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Yeah, I'd boost 2 or 2.5 to 3.5kHz about 3db and also possibly raise the treble region a bit, which is weird since I thought I prefer more upper midrange and more treble than you do. At least based on the HD800 and Ether.
    A lot of them also still seem to consistently have the same peaks and dips, but I don't think you can fully get rid of that without making other graphs weird. I guess that's always a compromise when you measure through ears, but something is still weird about the graphs. Maybe it's just that the materials and dimensions of the head are very different than the real thing. For example why does the Ether have a 5db boost sub 1kHz?
     
  15. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    I actually think this is reflective of the OG Ether's frequency response. I've always said there is something cloudy or about Ether's bass despite good distortion numbers. It's funny because I have all these headphones on hand and asked my wife to try them all, and she said the exact same thing about the Ethers.

    If you look at it as if Ether has a boost from 80 to 400Hz, the plateau, then it makes sense. Also, the Ether does sound muted from the high mids to lower treble. Not as much as any iteration of the LCD series though.

    As far as the screwy measurement behavior, you are exactly right. Fixing one thing for one headphone messes up something for another. I just had to work out a compromise.

    Finally, I'm loath to bring up treble on the plots more than I have already for fear of making it seem the HD58X may be bright sounding or having tizzy 10kHz emphasis which it does not.
     
    Last edited: Oct 7, 2018
    Serious and ultrabike like this.
  16. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    65,479
    Dislikes Received:
    108
    Trophy Points:
    113
    Location:
    Antarctica
    ^ Done.

    I think I'm finally done with these. Indicators of above changes are ESP95X and HD800, which don't look depressed here, and HD660X, which isn't. Pretty much as good as EARS can get for an approximate perceptual target as neutral as straight line across. I'm running with these.

    comps3.png [/QUOTE]
     
    Josh83, Walderstorn, Serious and 3 others like this.
  17. Vtory

    Vtory Illogical Spock

    Friend MZR
    Joined:
    Sep 12, 2016
    Likes Received:
    6,046
    Dislikes Received:
    38
    Trophy Points:
    113
    Location:
    Greater Philadelphia Area, PA
    Worked on developing a good compensation curve for myself (and my EARS) for one month. Managed to come up with pretty acceptable results after two big major changes. Below are the results of some famous headphones that I believe to serve as reference data points.

    More measurements to be uploaded during Nov, including Auteur (loaner), Ether 2 and WH-1000XM3.

    upload_2018-11-5_18-55-40.png
     
    Lyander likes this.
  18. monacelli

    monacelli Friend

    Friend
    Joined:
    Jul 23, 2017
    Likes Received:
    1,437
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    The Land of Enchantment
    Nice work, dude. If I were going to nitpick, I would ask if the HD600 treble sounds like that to you north of 10 kHz? They don't sound to me like they have a 10 dB peak near 17 kHz. This EARS system is cool in a "power to the people" kind of way. I would much rather have an informed individual taking measurements on this system than somebody with questionable motivations taking measurements on a mega-buck GRAS.
     
    Vtory likes this.
  19. Vtory

    Vtory Illogical Spock

    Friend MZR
    Joined:
    Sep 12, 2016
    Likes Received:
    6,046
    Dislikes Received:
    38
    Trophy Points:
    113
    Location:
    Greater Philadelphia Area, PA
    Your point is very valid. Honestly, I don't believe my result from 10khz, because that freq range is very subjective to subtle directional changes and positioning. I never thought a significant peak existed around 17k in hd600. But I don't believe Marv's result for Stax lambda from 15k hz (clearly to my ears there was more extension than measured), either. Furthermore I have never been assured by measurements from BK4128 and GRAS for that freq range.. haha. So, that's very difficult problem to me.

    For this reason, I just didn't touch anything from 10k hz. Minidsp corrected somewhat.. but still questionable. I definitely need more data points to adjust that super-high frequency.

    And also agree that EARS opens the door of measurements to more people. It's a big pleasure and fun that I can take any arbitrary measurements whenever I need. For example, I am looking at my measurements regarding reverberation times, clarity measures, and step responses... there were lots of interesting points to think of -- potentially meaningful time-domain characteristics.
     
    Last edited: Nov 5, 2018
    monacelli likes this.
  20. Walderstorn

    Walderstorn Friend

    Friend
    Joined:
    Mar 20, 2016
    Likes Received:
    893
    Dislikes Received:
    4
    Trophy Points:
    93
    Location:
    Europe
    Just ordered mine but i never measured anything. Should be here in 5 or 6 days. It's more of a "get-to-know" and having a little fun experience than anything else.
    I do have a question regarding IEM's though, if the canal depth can be a hinder ( i saw your post @james444) and since i usually use foam (comply) because of my very small ear canals, that won't even accept most "S", would that still be an issue?

    I will ask for help with the compensation afterwards, if anyone is willing to help a less savy member.
     

Share This Page