CSDs and What They Tell Us

Discussion in 'Headphone Measurements' started by purr1n, Sep 28, 2015.

  1. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    But interestingly speaking, I find them pretty nice for some music styles. Some metal bands are pretty good with these, it changes the sound clearly, but a lot of fun :).
     
  2. srkram

    srkram New

    Joined:
    Apr 16, 2017
    Likes Received:
    12
    Trophy Points:
    3
    Location:
    UK
    How about the effect of ringing below 2k? This is where you typically see the most, or the longest, ringing in the CSDs. Will ringing in the bass region cause the bass to sound muddy and slow?

    -Mark.
     
  3. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The area from roughy 750Hz to 1500 tends to show effect of pads and enclosure. The more open headphones have less "stuff" here. Below that, I don't think it means much with headphones because we are talking about wavelengths the size of tables, rooms, and auditoriums.
     
  4. Maxx134

    Maxx134 Dunning–Kruger effect poster boy

    Pyrate
    Joined:
    Oct 27, 2015
    Likes Received:
    818
    Trophy Points:
    93
    Location:
    NYC
    I have been eye straining for quite a while looking at wiggly lines with these CSD,
    and have to say this thread is so valuable to me for this!
    Saved me from hair pulling and blindness lol, thanks Marvey and all here.

    Also,
    I have noticed the more apart the lines are, generally mean more speed of decay, correct?
     
  5. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'll continue this conversation on other uses of the CSDs. Mainly want to do this because an armchair QB on reddit dismissed CSDs as being useless since we already had FR. To me, CSDs provide supplementary information to FR. The reality is that the FR and CSD are both mathematically derived from the impulse response. Sometimes different ways of looking at things allow us to glean different kinds of information.
     
  6. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Lesson 1
    • CSDs help us determine whether FR nulls are truly nulls or actually hidden peaks in disguise.
    • Longer lived ridges in CSDs sound worse than short lived ridges.

    So here's the FR of the Denon AH-D7000 (this is the V1 where I used a porous plate, so bass might be less with current measurements):
    [​IMG]

    And here's the FR of the Beyer Amiron:
    [​IMG]

    From the looks of it, we might think the Amiron has more difficult treble than the Denon D7200. Sure the Denon D7200 response looks more ragged, but its overall treble repsonse, despite a few peaks here and there, isn't that much higher than the mids from 500-1.5kHz. However, I subjectively I found the D7200's treble more difficult to bear (it was also more spotlit) than the Amiron's treble, despite the big middle finger at 8kHz in the Amiron's FR. This is one area where CSDs become useful and provide us with information above and beyond the FR.

    As we can see in the D7200's CSD, that sharp null in the FR plot* is actually nasty treble ringing at ~8.5kHz. Note that the ridge stays strong almost up to 2ms, and takes its time decaying up to 4ms.
    [​IMG]

    Now here is the Amiron CSD. Yeah we see this huge peak at 8kHz, but we also notice there is no long lived ridge. The decay of the peak is dead by 0.75ms. Now I'm not saying this is pleasant, because it isn't. But I did find the Amiron's treble subjectively easier to deal with than the D7200's. It was certainly bright, but the treble sounded smoother and less piercing than the D7200's treble.
    [​IMG]

    *Sharp nulls that happen to be peaks in disguise also happen with speaker measurements. They seem to occur with more regularity on headphone measurements, possibly because the internal cavity between the ear and the earcup produces a jumble of reflections and standing waves.
     
    Last edited: Sep 20, 2017
  7. Melvillian

    Melvillian Friend

    Pyrate
    Joined:
    Jul 3, 2017
    Likes Received:
    1,345
    Trophy Points:
    93
    Location:
    Los Angeles
    For the D7200 is that resonance at 1K that’s present until 4ms?
     
  8. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Tell the M1060 that CSDs are useless, and maybe it will sound less like a tuning fork.
     
  9. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    That will be covered Lesson 2.
     
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Well that would fall in line with Raqqot's (Redditor Armchair Quarterback who Quotes Olive and Toole) assertion that is all we need is frequency response, since the FR of the M1060 shows a sharp huge spike where that nasty ringing resides. However, in this the case, the CSD does serve as another handy visualization to make things that much more obvious for us.

    Again, the CSD and FR are both arrived at the impulse response - IR. We could of course say we don't need to look at the FR or CSD, and all that we need is is the IR (super zoomed in and across to a 100ms or so). but the fact is, FR and CSD are different presentations that make things much easier for us humans to digest. Different ways of looking at the same thing isn't bad or redundant.
     
    Last edited: Sep 20, 2017
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Here is another example related to to the Sony MDR-Z1R. Some of you may remember a few months ago yet another JudeGate incident where Jude claimed that his Z1R measurements didn't have a sharp 10kHz peak (like Tyll's did), and then proceeded to present e-peen by bragging about his high-end measurement gear.

    Several other homemade measurements confirmed the sharp 10kHz peak from Tyll's measurements. But Jude basically said his measurement rig did not show this 10kHz peak and that he did not hear it (personally I think he's deaf). My own Z1R FR measurements were not totally conclusive.


    Here we can see the Z1R has a bump in the mid-treble area, not much higher in level in relation to the midrange; but there's no evidence of a sharp peak.
    z1r-r.png

    That is not until we examine the CSD, where we notice that the tiny little notch just above 10kHz is actually a peak. Note that ridge just above 10kHz that extends up to about 1.4ms
    Z1R R.jpg

    Here's an overhead look. Another way to look at things that might make interesting stuff more obvious.
    Z1R R.jpg

    And I haven't even mentioned that peak / ridge at 3kHz yet! I wonder why Jude didn't hear that (he conveniently ignored it) because his measurements have that exact same frequency response peak.

    So really, the CSDs are supplementary to frequency response plots. Most of the time, we probably don't need them. But sometimes they come in handy providing us with extra information.
     
  12. Thad E Ginathom

    Thad E Ginathom Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    14,135
    Trophy Points:
    113
    Location:
    India
    This is really useful and presented in an easy to grasp way.

    Joining just now, I had several questions: luckily I went to the start of the thread before asking, and found them, and many more, answered already :)
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Too hard to say anything pertaining to the bass for CSDs. Bass wavelengths are SUPER long. For example, 20Hz is 50ms! The CSD plots above only show up to 3 to 5ms. At at these 50ms / 17meter wavelengths, CSDs for bass will be more indicate of room interactions. Another factor is how easy for bass measurements are polluted by environmental noise. 50ms is a long time to capture environmental pollution! And there tends to be a lot of extraneous bass noise in urban or suburban environments (cars, doors, planes, etc.). Muddy and slow bass is more easily the result of too much bass and/or high distortion. Maybe the HT guys have more insight into CSDs of the bass region; but for headphones, I haven't studied it enough because of the limitations I stated above. Maybe if I ever get a 1000lbs metal isolation chamber like Jude, I might take a shot at it.

    Lesson 2
    The 700-1.5kHz region might be interesting in terms of giving us a general sense of the effect of pads, internal cup reverb, etc. which is the next topic of discussion.

    Basically, the more stuff in this area, the more reverby or internal cup cavity effect we hear. Let me provide four examples. HD800, Z1R, HD650, and Atticus. of these five headphones, I would rank their them in this order from reverby to open.

    1. Atticus: closed headphone, nice reverb effect, or maybe not so nice depending upon your preference.
    2. Z1R: closed headphone, but Sony did a great job here of controlling the internal reflections, with their speshal cup design and all.
    3. HD650: open headphone, pretty darn open.
    4. HD800: probably the gold standard in open sound.
    Now let's see how the CSDs play out, paying attention to the region from roughly 700-1.5kHz. The CSDs below have lengthened time scales to 5ms.

    Atticus L.jpg
    Z1R L.jpg
    hd650 R stock.jpg
    HD800 stock L.jpg

    Do we see pattern here corresponds to how "open" or how "reverby" a headphone might sound? Yup. I'm sure a few Raqqots right now are furiously digging through Google searches finding quotes from Toole or Olive saying that I am wrong, and that this is all coincidence. I'll let you guys decide.

     
    Last edited: Sep 20, 2017
  14. spwath

    spwath Hijinks master cum laudle

    Pyrate BWC
    Joined:
    Dec 13, 2015
    Likes Received:
    7,894
    Trophy Points:
    113
    Location:
    Madison, WI
    Is there a lesson or thoughts on how to properly make a CSD, ie the right settings and stuff for the chart? I never know what to do with that.

    I liked the MDR-Z1R, I did not really notice any 10k peak, but I certainly measured one. And I'm still new to this audio thing, and heard the Z1R a few months ago too.
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The 10kHz peak is relatively minor. Per the CSD above, the ridge dies at 1.4ms which is more than 1ms, but not that much more. Also 10kHz tends to be much more benign than say a 8kHz peak. 1ms is kind of the threshold for me. Other factors are amplitude of peak, and the spot where the peak is at. There are personal sensitivities too. @Hands seems super sensitive to 3-4kHz; 5kHz peak doesn't bug me (maybe I have hearing damage here); I can't stand 7kHz, etc.

    As far as CSD parameters, I hard coded certain ones such as rise time and shape to give me the visual results that I wanted for the timescale (3ms to 5ms) after trying out a few different ones. The rise time of the FFT window that I use is super fast, just a few samples. What's important is to place the start of the window at the right spot. I wrote an algorithm in my CSD processor to detect where the peak of the impulse response was, and then go back two samples for the start of the rectangular portion (post rise) of the FFT window. (I think, I can't remember what I wrote in the code, the code is a mess. and I didn't document it very well).
     
    Last edited: Sep 20, 2017
  16. Melvillian

    Melvillian Friend

    Pyrate
    Joined:
    Jul 3, 2017
    Likes Received:
    1,345
    Trophy Points:
    93
    Location:
    Los Angeles
    Definitely heard the resonance in the Eikon, which prompted me to ask about resonance earlier in the thread. It’s a strange effect (nice for rock) because the headphone is not slow, but acoustic instruments still sound a bit soft compared to the open HD650. Pretty cool being able to actually see that in the CDS.
     
    Last edited: Sep 20, 2017
  17. Deep Funk

    Deep Funk Deep thoughts - Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    9,028
    Trophy Points:
    113
    Location:
    Amsterdam
    Home Page:
    Can you get the DT48 to work? If that thing can be properly measured you find some weird things with that driver.
     
  18. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I don't think it always works like that. Your LCD-X CSDs looked fantastic, mainly because the FR was so flat below 1kHz (low phase shift = clean CSD). Same thing for the UE4 CSDs. Both of these don't sound very open to me. Same thing with my RE-400 measurements: Fantastic CSDs below 2kHz, very closed-in sounding.
    I think it's more a matter of general FR smoothness, lack of bumps/bass emphasis and bass extension. In your example the Atticus measurement has quite a few tiny bumps between 200Hz and 1kHz, which manifest in the CSDs. And the better the measured bass extension, the cleaner the CSDs will look: RE 400 measures with better bass extension than UERM, gets cleaner CSDs down low and yet it sounds less open to me.
    But I think I get what you mean. Wonky resonance effects will usually manifest somehow in the FR, which again can easily be seen in the CSDs.

    For reference:
    LCD-X L
    [​IMG]

    UE4 L
    ue4l csd.PNG

    My own RE-400 (modded) measurements:
    RE 400 CSD.png
     
  19. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,779
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Here are the LCD-X measurements again (right channel - I can seem the find the left channel measurement) when used with the same parameters (gating, floor, timescale)
    LCD-X CSD.png
    LCD-X CSD2.png

    I don't provide CSDs for my IEMs measurements (for other reasons); hence my assertions were never meant to apply to IEMs. Nevertheless, regarding the IEM examples: they may not represent an open sound, but they don't exhibit internal cup effects - which is clearly indicated.

    Finally, I'm sure there may be exceptions to the rule, but the trend is hard to deny.
     
    Last edited: Sep 21, 2017
  20. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,408
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    Rreally digging the top-down view of CSD. Please include those in future measurements if you can be bothered!
     

Share This Page