Search This Blog

Friday, May 24, 2013

A new criterion will be added to the dataset.


Finally, Dr.Olive has come up with a headphone response target, which is subjectively verified by expert listeners. As I have suspected, it's more like a conventional diffuse-field target with a downward slope of -1 dB / octave within 100 Hz to 10,000 Hz. (Please refer to the actual paper for further detail)

And here's an assortment of various IEM's frequency response compensated by utilizing the Olive-Welti target. Pretty dramatic.

And of course, my modified Sony MH1 accurately follows the target! I might go ahead and attenuate the treble just a little bit here..

54 comments:

  1. Not much different from the Etymotic target curve.

    ReplyDelete
    Replies
    1. They are quite different: about -2.5dB at 1 kHz, and anything above 3 Khz, -5 dB. This difference ultimately makes ER4 way too bright, according to the target.

      Delete
    2. Sure, I meant it looks more like it's in the same family as the Ety curve compared to other DF-curves.

      Delete
  2. Are you going to be updating the other graphs?

    ReplyDelete
    Replies
    1. Nah, this will simply replace my "recommendation" section at the end of the article, as the conventional idea of diffuse-field oriented localization still stands.

      Delete
  3. The Olive-Welti target curve is not -2db/oct. It is too tilted target. I think that it is more like -1db/oct which is similar to the room target suggested by Roger Russell. (See

    Loudspeaker Measurements and Their Relationship to Listener Preferences: Part 1 by Floyd E. Toole, 1986.

    )

    P.S. The Ety's target is significantly different to Sean and Welti's target, but the magnitude response of ER-4P well follows it without bass range. Thus, I think that ER-4P gets to have good spectral balance with some bass boost. :-)

    See https://dl.dropboxusercontent.com/u/91668521/er-4p_response.png

    (The response of ER-4P was obtained from your magnitude sheet of in this post http://rinchoi.blogspot.kr/2012/03/etymotic-research-er-4b.html and the difference of responses between ER-4P and ER-4B from the electric properties of these two earphones.)

    ReplyDelete
    Replies
    1. Well spoken, Carrot! It is indeed -1 dB/oct. It seems the slope is somewhere between Moller and Russell, as the angle gently shifts at 100 Hz. Nonetheless, we get the idea here that the current model is far from the originally claimed raw-flat, the independent-of-direction target, or Golderears' pseudo-scientific God-forbidden reference bullshit.

      Delete
    2. (In Korean)

      예전에 골든이어스 타겟과 논문의 RR 정도의 타겟을 비교해봤었는데, 규모 자체는 꽤 유사하기는 했습니다.

      http://ko.goldenears.net/board/2261228

      현재는 해당 타겟이 좀 수정된 것으로 알고 있는데, 예전에 확인했던 걸로는 헤드폰은 4.5dB로 수정되고, 헤드폰/이어폰 모두 피크의 대역폭이 좀 커졌습니다. (Accudio에 반영된 사항)

      뭐 규모는 그럭저럭 괜찮다고 할지라도 그 보상의 근거가 참 불명확한지라 되게 애매합니다만... (사실 저역보다 고역의 보정이 훨씬 문제라고 생각함다.)

      Delete
    3. Maybe the ER-4S would match the curve even better - wasn't it designed to be diffuse-field compensated but with a high-frequency attenuation between 2 kHz ~ 20 kHz? It says here http://rinchoi.blogspot.co.uk/2012/03/personal-concerns-regarding-other.html that it matches the X-curve of a small room environment. What do you think?

      Delete
    4. As I said, the deviation between the Olive-Welti target and Etymotic Research's diffuse-field reference is about 2.5dB at 1 kHz, and anything above 3 Khz, 5 dB. No products from Etymotic Research match the Olive-welti target, not even close. :p

      Delete
  4. So considering the Olive-Welti is somewhere in between Moller and Russell that make the Ety and THX curve (w. flat bass and steeper slope starting in the mids) the elephant in the room, doesnt it? http://ko.goldenears.net/board/files/attach/images/3040/057/194/002/09506c36c117bf640f275612d1108696.png

    ReplyDelete
    Replies
    1. I think we need to match our terminologies here.. Please check this out first: https://twitter.com/arthurinchoi/status/338610601695772672/photo/1

      The one I am referring to is ER's original reference target, of which ER-1 and ER-4B are based on. The one you are referring to is ER's modified target, with -5 dB @ 10 kHz from the original, of which ER-4S is based on. Even with the modified reference, the mid-range level must be attenuated around -5 dB centering at 2 kHz with low Q to at least loosely match the Olive-Welti target. IOW, no elephant in the room, maybe a hippo? :p

      Delete
  5. I mean the Ety mofified target, THX curve, Olive-Welti, Mueller and Rusell are independent studies that should define, in this context, the same thing, yet there's a divide between the first two and last two. I guess it's down to the room used as a reference for each study.

    ReplyDelete
    Replies
    1. Small correction, the Mueller and Russel curve require you provide the DF-curve along with it.

      Delete
    2. Yes! The difference between each studies is the condition of the room used as a reference. And Harman's reference room is one of the best, if not the best, you can find within the industry. The reference used by Etymotic Research is awfully off in that regard? http://i.imgur.com/F6ukFsD.gif

      Delete
    3. http://seanolive.blogspot.com/2009/05/harman-international-reference.html http://seanolive.blogspot.com/2009/05/harmans-how-to-listen-new-listener.html
      Studio control rooms have bass traps in each corner and absorbers at the first reflection points. AFAICT Harmans listening room has acoustic treatment applied to none of these critical points. OTOH This better reflects the average listening room.

      Delete
    4. We want the same end result, regardless of what the environmental variables are in order to prevent "circle of confusion", which is defined by F.Toole, to achieve Hi-Fidelity.

      Furthermore, we can't possibly define a single room target of "studio control" rooms, as they even have their own linear distortion to be taken care of: http://2.bp.blogspot.com/_w5OVFV2Gsos/Su3b8m6wHjI/AAAAAAAAALw/ltcrAglRR4g/s400/Makivirta+and+Anet+2001.png

      Is Etymotic Research's target really off then? Here's what Mr.Don Wilson, the designer of ER-4S has to say about the modified room target: "Several engineers and myself did a study in a radio station recording studio in Chicago for an AES meeting when the ER-4's first came out. We compared several different recordings through a pair of room equalized loudspeakers. The participants listened to the same recording through the loudspeakers and the ER-4 one ear at a time. We played the recordings through the ER-4B with several different equalizations. The participants voted as to which response sounded the closest to the loudspeakers. After we graphed the answers we discovered that the ER-4S had the closest response to the speakers.."

      As you can tell, Etymotic Research's modified target is designed without having a proper protocol for subjective assessment. Which means, it has no scientific significance.

      Delete
  6. Yes, I'm familiar with Olives study of studio control rooms. Regarding Etymotic target even though the protocol wasnt rigourous, they ended up on a result that corresponded with THX.

    My point is Olive-Welti, Russel and Mueller are together working toward preventing the circle of confusion, like you say. Etymotic and THX are doing the same in a parallel universe! An exageration, of course, but I hope this gets my point across, please let me know if you think I'm unjustified in this assertion.

    One question, in the 'Deviation of Etymotic Research's references' graph you posted earlier, by 'reference' do you mean the blue line or the dashed red line here: http://www.etymotic.com/ephp/images/er4_graph_wrfy2.png

    ReplyDelete
    Replies
    1. Nevermind the question, your latest tweet confirmed my suspicion.

      Earlier today I went ahead and created a filter from your Olive-Ety deviation graph. It sounded too murky for my taste but with the magnitude scaled to 50%, and in the same chain a filter that corrects to the ER4S reference, also magnitude scaled to 50%, the combination sounds very nice to me. I have to admit you were right, the ER4S is a bit excessive between 1-3k.

      Delete
    2. I am glad it worked out! ER's modified target was indeed a hippo ;) Maybe we can come up with a passive filter, which turns ER-4 into an ER-4Olive-Welti, LMAO

      Delete
    3. Maybe so. MAutoEqualizer is a piece of software I use to generate equivalent parametric eq values from a continous curve.

      Delete
    4. Simpsons did it!
      http://www.linkwitzlab.com/reference_earphones.htm
      Joking apart, I always felt er4s too harsh on my ears and following equalization attempt of Linkwitz actually makes er4s quite listenable. My second peek is around 8.5 khz, not 7.5 though.

      Delete
    5. Yes, but since Linkwitz published that article we have additional data to quantify the responce by objective means.

      About your your 8.5k peak:
      http://rinchoi.blogspot.se/2010/05/how-deep-do-you-insert-you-er4.html

      Delete
    6. Mr. Linkwitz designed the filter based on his assumption that the second wavelength resonance always resides at 7.5 kHz, which is equivalent to the insertion depth roughly 10 mm away from the reference plane. Here's the result of my quick simulation: http://twitter.com/arthurinchoi/status/339489083611025409/photo/1

      The filter should work, as long as the insertion depth is matched at the same position every time the IEM is inserted, but I would just insert ER4 deeper, or just get a different IEM designed for shallow insertion as his filter is quite messy & lossy.

      Delete
    7. Guys, really help me to get er4s deeper into the ear. I am bound to gliders, as silicone flanges just hurt too much. ANd i tried both frost ones and normal size as well.

      Delete
    8. Small update: My filter preference eventually shifted to 100% ER4S reference + 50% Olive-Welti. Never enjoyed my ER4S as much as I do now. I can't quantify how much insertion depth is affecting my subjective asesment however. Here's what my molds look like: http://i.imgur.com/4wXmxhN.jpg

      Delete
    9. Vitaly, you could try some third party tips, such as Shure Olives. Or you could take advantage of the custom-fit program. In such case make sure the lab gets this note: http://www.head-fi.org/t/538615/if-you-still-love-etymotic-er4-this-is-the-thread-for-you/1275#post_9402331

      Delete
    10. Ah, maybe this is a good chance to present some analysis on ER's Custom Fit program, with/without damper location adjustment. I have total 4 pairs of the Custom Fit sleeves, with the former two made by Canadian DST Swiss and latter two made by ACS UK. I'll upload the data soon.

      Delete
    11. Thanks for suggestions, Mark.
      I am more inclined to try very small tips than using custom molds. I live on a rock in the middle of a sea and audiologists here are say "rare".
      So i was wondering which tips are the smallest and softest ones... Shure olives?

      Delete
    12. I've had some luck with small size Shure Olives.

      Delete
  7. I am not sure whether their methodology was correct though, to come up with an universal reference target. The problem here is, the spectral characteristic of THX theaters can't possibly be random-incidental, or diffuse-field oriented. I wonder how they came up with the data in the first place..

    Or we can simply assume the THX curve was derived from doing a simple room calibration, and subtracted the data from the ER's reference. Kinda makes sense. However, I still don't see much of parallel behavior here.(maybe a hippo? :p) Deviation from ER's modified reference to the THX reference, graphed: https://twitter.com/arthurinchoi/status/339186226143842304/photo/1
    I truly hope THX theaters don't introduce a wide plunge centered at 8 kHz.

    Meanwhile, I always refer to the original dashed one when I mention ER's reference target, please check this out: https://twitter.com/arthurinchoi/status/339174252437700609/photo/1

    ReplyDelete
    Replies
    1. You must mean Ety-DF AKA ER4B reference in the first graph. I agree it seems odd that THX would use a 8k cut after room calibration. I'll concede your point for now.

      All I can add is that pro audio studio moniors historically had more 1-3k, probably for intelligebility reasons, compared to hifi speakers. The trend has shifted to a flatter frequency responce. Maybe Etymotic is following this custom in some way. But I don't feel like exploring this aspect any further and open another can of worms.

      Sorry the hippo reference is lost on me.

      Delete
    2. Hippos and elephants, my friend!

      Ooops my bad, I fixed the problem here, LOL I should stop eating and graphing at the same time.. https://twitter.com/arthurinchoi/status/339211155929182208/photo/1

      Regardless, they are nowhere close to each other :( Perhaps Etymotic Research is still stuck with the classical notion of recording-playback relationship, considering the fact that Dr. Killion's main area of expertise is hearing aids acoustics, not Hi-fidelity.

      Delete
  8. What you guys talking about is beyond me. But, if i'm not wrong, the modified fq graph of sennheiser ie800 is flatter than i had expected except an obvious dip in 3khz. So ie800's bass is not that exaggerated. right?

    ReplyDelete
    Replies
    1. That would be the right assumption.

      Delete
  9. So, with the damper correction EQ in place my preference is closer to Olive-Welti, but not 100%, that still sounds scooped and murky to me. Right now I like 70/30 Olive-Welti/ER4S reference.

    ReplyDelete
    Replies
    1. It's like I'm finally getting the most out of my ER4! It's thanks to your dedication. Hats off to you, Rin!

      Delete
    2. I am humbled, my dear friend. Feel free to let me know if you have any concern!

      Delete
  10. Am I doing this right? With the ER4S and small triflanges(maybe 1-2mm short of ref plane) I listen back with filters generated from http://www.etymotic.com/ephp/images/er4_graph_wrfy2.png black line inverted, followed by https://pbs.twimg.com/media/BLL8jFfCAAANeiG.png. Sounds weird until I bypass the first filter.

    ReplyDelete
    Replies
    1. I'd just use the green curve, as you are not sure that your in-ear response is equal to that of ER's Zwislocki coupler. Just my 2 cents.

      Delete
  11. A Whole Lot of Hooey...

    I prefer raw data. Then if Y'All want to add someone else's thoughts for a compensated room why not show them all? And, Not just your favorite one?

    Two Cents from the Dumb Guy, Jim

    ReplyDelete
    Replies
    1. That sounds great, Jim. I wish I had kept the QSC Audio's ABX comparator LOL so that we can do our own experiment as well.. :X Well, some data are scientifically more robust than the others, under a specific circumstance/criterion- and I would like to honor their significance. Can't do all of them, but at least the few I am aware of.

      Delete
  12. Rin,
    I understand what you are doing here, your blog, and thank you for it.
    I would rather not see you swept up with the flavor of the day though.
    Whether it be your friends or someone elses compensations... They are all subject to change.
    But the raw data will live on without any compensations applied.
    If there are multiple compensation theories, shouldn't they all be shown?
    What is to say that the valued opinions of Olive won't change?

    Still trying to get a handle on all the stuff, Jim

    ReplyDelete
    Replies
    1. "the raw data will live on without any compensations applied"

      The Statement of the Year, Jim! The 'modified RR' target that I am currently utilizing is indeed quite preliminary, as it is basically a spectral sum of Harman's quasi-HATS system in a reference room. I am simply using the target, because it certainly marks the beginning of the new era of headphones, jump-starting all the fuss about "the perfect headphones" and so forth.

      More claims from different parties will arise in the future, emphasizing their own scientific significance. Regardless, the conventional diffuse-field target will always remain as my reference, as it is the correct representation of the resulting orthotelephonic gain(localization characteristic) in accordance to the association model, whether the perceived tonality ends up being bright or not.

      Delete
    2. Rin,
      Thanks for your reply...
      Perhaps more important to the end user/consumer than the headphones ability to give good raw data, is it's ability to respond well to equalization/end user adjustments.
      You see, I am of the opinion that if I can buy a good quality headphone that will respond well to some tweaks along the way to my music enjoyment, Then I have fulfilled my true goal.
      Musical Enjoyment.
      That Info, along with different effects of peripheral equipment are some of the basic answers that I need as a consumer.
      I am rambling now...
      Jim

      PS Find a Ten'r in the pot for some liquid refreshment. Enjoy the 4th!!!

      Delete
    3. Absolutely, musical enjoyment is the key, and is the sole reason I am running this blog. What good is a headphone, if there is no music? :D

      I'll do my best to keep the spirit alive and to the proper reference, thnx a lot for all the contributions!

      Delete
  13. I hope I'm not being a nuisance by asking again but I would be very grateful if you could find the time to make a Olive-Welti deviation graph for the Sony MDR-7506 based on GE's graph:
    http://en.goldenears.net/4726
    I'm going on the assumption that you have an idea of the compensation curve GE used, let me know if this is not the case.

    ReplyDelete
    Replies
    1. Oh, not at all Mark! I just need to get the headphone analyzed first though.

      Delete
    2. Ok, thanks for letting me know.

      Delete
  14. looking at all these comments, have either of you tried the er4s with red dampers? it lowers the 1-3khz area in a way and some of the whole treble area making it more sloped towards the bass end a tad. how would that compare to the targets you are commenting about?

    ReplyDelete