Page 1 of 2 12 LastLast
Results 1 to 10 of 13

Thread: Subjective Evaluation of Distortion Performance

  1. #1
    Devil's Advocate Adam_MSS's Avatar
    Join Date
    Aug 2010
    Posts
    1,578
    Real Name
    Adam

    Subjective Evaluation of Distortion Performance

    Quote Originally Posted by winslow View Post
    About distortion...is there any links or evidence to support that low distortion equates to any sublective performance benefits?

    Then again you might pick a distortion profile because you like a certain sound...which is something that I want to learn how to analyze.

    A bit of a complex subject, but I think a reasonable place to start might be in discussing some literature.

    Geddes published a series of papers and presentations on his own distortion metric. He was investigating a new metric precisely because the existing default metric (THD) was often not closely correlated with listener preference.

    In Geddes work, he looked at the auditory perception of distortion and uses that to set the principles of the metric:

    http://www.gedlee.com/downloads/Distortion_AES_I.pdf
    Given these characteristics we will propose the
    following three Distortion Perception Principles.
    1. Distortion by-products that are created
    upward in frequency are likely to be less
    perceptible (masked to a greater extent) than
    those that fall lower in frequency.
    2. Distortion by-products that lie closer to the
    excitation are less likely to be perceived
    (they are masked) than those that lie farther
    away (masking is a localized effect it only
    occurs in the vicinity of the masker).
    3. Distortion by-products of any kind are likely
    to be more perceptible at lower signal levels
    than at higher signal levels.
    So in the GedLee metric, various distortion products are weighted differently based on their audibility and impact on perception. Higher odd-order harmonics being particularly bad, while something like 2nd order distortion might not be as bad, or may even be perceived as pleasant (warmth!).

    In the second paper he shows the correlation of his metric to listener preference and also the correlation of THD and IMD. It's a compelling argument.

    We can put it into practice when looking at distortion plots for various drivers/systems and using the principles to evaluate the plots. We could then for example prioritize elimination of higher order odd harmonics, even if it means accepting some lower even order distortion.

    This is just the tip of the iceburg as I'm largely talking about driver distortion. There are other sources of distortion upstream that can be issues as well, and Geddes discusses them to some extent in the papers. (He actually has a testing procedure for amps and other electronics that appears to be very effective based on listener accounts. It's how he chose his $100 receiver to drive his demo loudspeakers.)

    Perception
    http://www.gedlee.com/downloads/Distortion_AES_I.pdf
    http://www.gedlee.com/downloads/Distortion_AES_II.pdf
    You don't use science to show that you're right, you use science to become right. - R.Munroe

    The important thing in science is not so much to obtain new facts as to discover new ways of thinking about them. - W.L.Bragg



  2. #2
    Senior Member
    Join Date
    Sep 2010
    Posts
    430
    I have (and did my best to read) Geddes's book...if anyone wants to borrow it they can. But be warned, the math is rather heavy and you really need to know Linear Algebra (my nearly under grad in math with a single LA class didn't help me any).

    I will re-read the papers....be a while since I last read them. I re-read Linkwitz's site last month....there is so much in there that I can't apply everything I've learned at once.

    But I figure speaker distortion would be as good of a place as any to get the ball rolling.

    I have my own beliefs, but curious as to where we can go with this stuff.

  3. #3
    Tester Extraordinaire ErinH's Avatar
    Join Date
    Aug 2010
    Posts
    1,247
    Quote Originally Posted by winslow View Post
    ...you really need to know Linear Algebra (my nearly under grad in math with a single LA class didn't help me any).
    Just a suggestion: if you have a copy of mathcad, it'll make easy work out of it.
    Your ears: The best tools you have... and they're free, too!

  4. #4
    Devil's Advocate Adam_MSS's Avatar
    Join Date
    Aug 2010
    Posts
    1,578
    Real Name
    Adam
    Speaking of mathcad, I need to dig around a bit, but I might have the mathcad filters/files that Geddes used to create his files with varying forms of distortion for some of his tests. I know I had them at some point but didn't have access to mathcad at the time.
    You don't use science to show that you're right, you use science to become right. - R.Munroe

    The important thing in science is not so much to obtain new facts as to discover new ways of thinking about them. - W.L.Bragg



  5. #5
    Senior Member
    Join Date
    Sep 2010
    Posts
    430
    No, we didn't use Mathcad...I have a lot of statistical software, but no mathcad.

    I knew I should have been an engineer.

  6. #6
    Founding Member OSN's Avatar
    Join Date
    Aug 2010
    Location
    Philly
    Posts
    252
    I'm one of those warmth-seeking guys. When I am evaluating drivers, I seek distortion profiles and the higher the even order distortion in relation to odd order distortion, the more likely I am going to like the way they sound. Then it's a matter of whether it will work with my enclosure space available, design goals, and power handling. I was very curious when I was asking myself 'why do I like these drivers so much?' and the distortion profile usually explains it. Paper cone drivers generally win with me.

  7. #7
    Devil's Advocate Adam_MSS's Avatar
    Join Date
    Aug 2010
    Posts
    1,578
    Real Name
    Adam
    The whole concept of low even order distortion being preferable is interesting to me. I've experienced both with loudspeaker drivers and with tube amplification.

    I'd really like to run some of my favorite audio tracks through mathcad to add in some 2nd and maybe 4th order distortion at varying levels to see if I have a preference. Obviously it's impossible to remove whatever distortion exists in the loudspeakers/electronics used for the test so it would be critical to use a very low distortion system for playback.

    Then it would be cool to have some sort of on-the-fly filter available to add in distortion at will. I'm sure the ability probably exists already. In particular I think about all the various guitar amplifiers that with various modeling functions.
    You don't use science to show that you're right, you use science to become right. - R.Munroe

    The important thing in science is not so much to obtain new facts as to discover new ways of thinking about them. - W.L.Bragg



  8. #8
    The concept of "weighting" distortions that are more and less objectionable is attractive in concept, but involves some contentious assumptions, I suspect. But every speaker system designer should be designing for acceptable weaknesses over unacceptable ones, right?

    I need to read up on Geddes' stuff, as the overviews that I've read seemed to contradict some of my experience.
    "Of course, the laws of science contain no matter and have no energy either and therefore do not exist except in people's minds. It's best to be completely scientific about the whole thing and refuse to believe in either ghosts or the laws of science. That way you're safe. That doesn't leave you very much to believe in, but that's scientific too."

  9. #9
    Senior Member
    Join Date
    Sep 2010
    Posts
    430
    I have been really thinking about this lately.

    I have noticed my tweeters are pretty "detailed" sounding. Then I go and look at distortion plots and see a rise in the 3rd order over the 2nd order around 6-8k. This has me thinking is the detail real or a product of distortion? I have experienced similar things in the past but haven't given much thought to it. I am still wondering if the distortion plots are audible in terms of amplitude or if they are below hearing thresholds.

  10. #10
    Founding Member 86mr2's Avatar
    Join Date
    Aug 2010
    Posts
    286
    Quote Originally Posted by winslow View Post
    I have been really thinking about this lately.

    I have noticed my tweeters are pretty "detailed" sounding. Then I go and look at distortion plots and see a rise in the 3rd order over the 2nd order around 6-8k. This has me thinking is the detail real or a product of distortion?
    That was Zaph's contention about ribbon tweeters. It was controversial in the early days of DIYMA as ribbons were one of the early foci of the community. It always made sense to me. There should be some old posts by Nguyen about the ribbon test and Zaph's comments over there yet.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •