Active Device (op-amp/tube) coloring and distortion discussion

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by Hrodulf, Oct 30, 2018.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Over time, I've come to appreciate the OPA2134. Audiophile stuff like the OPA627 sound colored however which way you cut it - anything with FETs running audio signal. JFETs, DiFETs, MOSFETS.

    ...

    Here's the deal. I have a philosophy and approach. It isn't static. It changes. But at least I have one, right or wrong. It works for me, but it may not work for any of you. Now the people who don't have any approach - they are the ones who tend to buy random shit, embark on the journey of countless side-grades, mismatch resolving gear with shit gear, or worry about how a line level transformer is going to have 1% distortion at 20Hz for a 22dBU signal when their self-powered studio monitors can't even do 40Hz.
     
    Last edited: Oct 31, 2018
  2. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    As far as my thoughts on "power adds", assume X and Y are two random signal sources. The total power of the two signals added together would be:

    E{(X+Y)^2} = E{X^2} + 2E{XY} + E{Y^2}

    where E{*} is the expected value.

    If X and Y are uncorrelated the cross correlations can be separated so that:

    E{(X+Y)^2} = E{X^2} + 2E{X}E{Y} + E{Y^2}

    If either E{X} and/or E{Y} is/are zero mean, we have:

    E{(X+Y)^2} = E{X^2} + E{Y^2}.

    Otherwise one needs to deal with the cross correlation term. One can verify this in Matlab or whatever.

    Note X can be source (music) signal and Y can be noise (as in noise floor). A long as the noise floor is zero mean and uncorrelated to the source signal, power adds.

    This cannot always be said about harmonic distortion (including intermodulation) which is correlated to the signal source. For those, power may not just add given cross-correlation terms.

    As far as transients, abrupt changes in frequency response tend to have long transients. The question is whether those transients are on the transition band edge, or somewhere in band.

    Deep nulls and narrow peaks in band are likely to generate transient like behavior in CSDs and those are difficult to fix. But I feel transients due to having a steep decline in the transition band shouldn't be much of a problem unless phase is all out of shape.
     
    Last edited: Oct 31, 2018
  3. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    ^ Doesn't this also explain why a 20Hz 0dFS signal is unlikely in any kind of content? Any add on to that would clip the signal - we'd run out of digital headroom.
     
  4. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    If one is out of headroom, one would clip. If one clips correlated non-linear distortion (to the 20 Hz signal) will show it's ugly head. I believe the power of a HF signal (riding on the 20 Hz) will add to the 20 Hz distortion if the HF is not harmonically related to the 20 Hz signal (frequency bins are orthogonal and therefore uncorrelated).

    One way or the other, shit will hit the fan. The frequency response should show some of that.
     
    Last edited: Oct 31, 2018
  5. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Yes. Op-Amps are active devices that can oscillate if not properly used.

    When not properly used, or if using the wrong part for the application, or in a poor design (even if not oscillating), results will likely be shit. But similar things could be said about plenty of active devices.
     
  6. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    The deal with the 20 Hz LF + HF signal is that power will add if the 20 Hz distortion harmonics do not occupy the same bins as the HF signal.

    Otherwise, things will not be that straight cut like @OJneg said.

    EDIT: Actually, also like you guys said, the HF will clip as well, and even more shit will happen.

    Two tones not clipping (40 Hz + 10050 Hz where LF tone is 20x the HF tone):

    two_tone_not_clipping.png

    Two tones clipping (quick and dirty asymmetrical clip to 10 arbitrary units on the + side):

    two_tone_clipping.png
     
    Last edited: Oct 31, 2018
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I obviously got lost in the weeds there with some technical details folks were alluding to. I see what you are saying.

    In another thread I remember you added some plots of actual content and many don't show much below 20 Hz (even 30 something Hz).

    Do recording studios specifically roll of < 20 Hz signal content to avoid these issues?

    All that said, I know a buddy of mine is crazy about sub-bass and wants to reproduce down to 16 Hz since he is big on organs. I personally couldn't care less to go down that low.
     
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Not necessarily specifically rolling-off bass, but turning down overall level to allow more headroom. So instead of that organ note peaking at 0dbFS / 32767 at 22Hz, maybe the engineers simply reduce the level of everything so that 22Hz is -6dbFS / 16000, to allow for other notes to pile on without causing clipping.

    I do wonder about the organ notes though and how much is fundamental or how much is harmonics. Low bass is more often felt than heard, and if it's heard, it's usually shitty harmonics. Real world sounds seem to be limited in energy below 35Hz, or perhaps more accurately, there if often more stuff going on above 35Hz than below it. Here is a spectrum analysis (max and mean) of that recent YouTube video I posted recording my car's exhaust. Microphone EQ correction was even applied to the bass. (High DR recording, will sound soft unless cranked up).

    charger.png
     

    Attached Files:

    Last edited: Oct 31, 2018
  9. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    Is anyone familiar with any vacuum tube opamps floating around out there. If we want to really pin down the differences in sound quality, maybe we should find a wholly tube based opamp (LTP differential input, ~100dB open loop gain, bandwidth compensated in typical fashion) and examine its perceived sonic characteristics. Otherwise we will continue to compare different designs as well as different devices.

    Given that single ended transistor designs are very scarce, I admit that classic tube amplifiers will still offer something that just isn't practical with solid state devices otherwise. Assuming the tube magic is lost when implemented in opamp fashion

    In fact, certain folks experience with classic SET designs that do use a certain amount of NFB might give us some clues as to what a tube design with a whole lots of NFB might yield.
     
  10. SSL

    SSL Friend

    Pyrate
    Joined:
    Nov 12, 2015
    Likes Received:
    1,239
    Trophy Points:
    93
    Mostly harmonics. I played organ in college; but this is true of most acoustic instruments anyway. Further, if an organ has multiple stops pulled then you'll get even more tones above the fundamental for a single note.
     
  11. Scott Kramer

    Scott Kramer Friend

    Pyrate
    Joined:
    May 3, 2016
    Likes Received:
    1,446
    Trophy Points:
    93
     
  12. maverickronin

    maverickronin Friend

    Pyrate
    Joined:
    Mar 17, 2018
    Likes Received:
    313
    Trophy Points:
    63
    Location:
    Midwest, USA
    IME they can help some, but don't completely fix it. Resonant modes still get excited even if there's nothing in the input at that frequency.
     
  13. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    That sounds like non-linear distortion. Could you provide a real life scenario?
     
  14. maverickronin

    maverickronin Friend

    Pyrate
    Joined:
    Mar 17, 2018
    Likes Received:
    313
    Trophy Points:
    63
    Location:
    Midwest, USA
    It is non linear. All the various HD800 mods for example.
     
  15. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    If you can solve a standing wave problem with a mod, I think that maybe better because it may not require the driver to work harder. Or maybe not. Proof would be on the pudding.

    But I think the stock HD800 response had good SINAD across all the passband. So PEQ and convolution should work well as long as it's not over done. And such things will have a positive and perhaps significant impact on decay.

    Personally, I don't bother with the HD800, and just use my HD600 w/o mods or equalization.
     
  16. maverickronin

    maverickronin Friend

    Pyrate
    Joined:
    Mar 17, 2018
    Likes Received:
    313
    Trophy Points:
    63
    Location:
    Midwest, USA
    I've never been able to solve that kind of problem with PEQ. Only make it less noticeable by lowering the level.

    Haven't tried static convolution since that's even more work than physical mods (making earbud mics...) and harder to use (even fewer sources allow for arbitrary convolution than have PEQ) but from first principles it really shouldn't be much better since it's just altering the frequency response, albeit more precisely.
     
  17. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The thing that scares me about convolving with IR is where you are sitting in the room. Bass is weird. It's different on different parts of the couch. I'm curious if "perfect" at the listening position might not be so when I lie down two feet to the left and two one foot down on the couch.
     
  18. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    It is possible to break an FIR (convolution) in second order cascaded sections and adjust the location of the complex zeros as needed similar to how it's done with PEQ. It may be a little different, but probably not much. However, I don't know of any application that supports that the way PEQ does.

    PEQ should work fine. It also depends of how many sections you have. I would say that 10 bi-quads should make a large positive impact on an HD800. As long as you are not allergic to EQ.

    EQ in audio is a one dimensional solution when dealing with a particular single driver. That said, something that is wrong at most listening angles can be addressed by a good EQ (PEQ or convolution) as long as there is sufficient SINAD.

    Bass can be hard. I can guarantee that making bass sound awesome in one location may screw things in another. Went through that exercise in my bedroom with the MiniDSP and the Datyon 10" sub. It seems having open baffle solutions help in making things more uniform across the whole room in the bass area. And if that is the case, EQ may work well.

    EDIT: In fact, Linkwitz solutions seems to rely on open baffle, wide radiation (omni or pseudo-omni response), and EQ heavily.
     
    Last edited: Nov 1, 2018
  19. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Cancellation of bass waves hitting the side of the room and back with dipole. IME, modes are still there, but halved in amplitude.

    Yes, I too have walked around the room with a mic to record bass response at different spots!
     
  20. frenchbat

    frenchbat Almost "Made"

    Joined:
    Sep 26, 2015
    Likes Received:
    1,369
    Trophy Points:
    93
    Don't think it matters when you're passed out drunk at the foot of the couch.
     

Share This Page