CSDs and What They Tell Us

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

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    CSDs and Ringing:

    9/20/2017 Update

    Lesson 1 | Above and Beyond Simple FR: http://www.superbestaudiofriends.or...types-of-driver-ringing.65/page-2#post-162886
    Lesson 2 | Cup Reverb vs. Open: http://www.superbestaudiofriends.or...types-of-driver-ringing.65/page-2#post-162951



    Cumulative spectral decay (CSD) or waterfall plots measure energy content over time across all frequencies based on a "sound burst" that excites all frequencies. They are sort of like a FR over time (technically not really as they measure energy throughout the band in successively smaller windows over time.) These plots are a good tool to "see" driver ringing or resonances (seen as ridges), which typically sounds bad (piercing, glare, shoutyness, etc.); and how fast and cleanly the driver decays (speed and extraction of low level information.)

    For a perfect or near-perfect response, the decay should be immediate. We first see a wall (meaning all frequencies are excited) at the beginning. The next few fractions of a millisecond, the wall collapses into silence (meaning the driver is no longer vibrating or sound waves are no longer bouncing around inside the headphone enclosures.) The back wall at time = 0 is basically an uncompensated frequency response. Successive plots over time move toward us.

    Here is quick guide to different types of driving ringing and how bad they sound to me. Your opinions may differ.
    • Ringing between 6kHz and 12kHz is bad.
    • Sharp and/or thin ridges, especially those that originate from clean floors, are nasty
    • Long duration ringing that doesn't stop by 2ms is worse than short duration ringing, which some exceptions, see [4] below.
    • Rounded ridges or bumps are more benign.
    • Ringing at 4-5kHz seems more benign than ringing below or above that range. Possibly because that frequency range tends to provide "edge."
    • Two or more resonances which are an octave apart, sound nasty, even if they are low in magnitude or short-lived.
    Ringing Regions:

    2-3kHz ringing:
    • Vocal shrillness
    • Guitar crunch
    • Shouty

    4-5kHz ringing:
    • Adds a bit of edge, bite, and excitement to the sound.
    • Detail (macro / coarse)

    6kHz ringing:
    • Generally tolerable, but unpleasant
    • Similar to 5k in that it adds some bite
    • Hardness bordering on glare
    7kHz ringing:
    • Glare with some sibilance
    • Sharpness

    8kHz ringing:
    • Very bad. Nasty. Painful.
    • Sibilance with some tizziness

    9k-12k ringing:
    • Adds definition
    • More tolerable, but still unpleasant if too much
    • Tizziness and sizzle
    • Grainy
    Ringing Case Studies:

    [1] The Sharp Lower Treble Razor of Death. The sharp razor of death that occurs anywhere from 6kHz to 11kHz will almost always be nasty. A sharp stabbing sibilant sound you will get. This is seen on D7000 and LCD2r1.
    [​IMG] [​IMG]

    [2] The Midrange Shout. A broad ridge between 1 and 3kHz. Tends to make things sound shouty or tinny. The HP1000 has a bit of this. So do all the Grados.
    [​IMG] [​IMG]

    [3] Upper Mid Ridge. Usually manifests as a slightly rounded ridge at 4-5kHz. To my ears, it's the most benign type and sometimes I don't even notice it, especially if the FR doesn't have a large corresponding bump. Examples are ATH-AD2000, ATH-M50, AKG-K240, HD558, and Grado RS-1 (the issues I have with the Grado are more with the thin ridge at 8kHz than the ringing at 5kHz).
    [​IMG] [​IMG]

    [4] The Double or Triple Octave Penetrator. Even if the ringing is not very severe, ringing or peaks an octave apart seem especially nasty to me. Below the K601 has ringing at 3.75, 7.5 and 15kHz. That a triple penetrator. The mystery headphone below has ringing at 5k, 10k, and 20k. The resonances are not all long lived nor of high amplitude, but their effect combined together is unpleasant.
    [​IMG]

    [​IMG]
     
    • Like Like x 15
    • Epic Epic x 1
    • List
    Last edited: Sep 20, 2017
  2. takato14

    takato14 God of Ruin

    Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    654
    Trophy Points:
    93
    Location:
    Michigan, USA
    "mystery headphone"

    thats the HD700 and you know it :p
     
  3. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Oh guys, I need serious help for this topic. I've heard several times this "ringing issue" ("heard", you get it ? :D) but I hadn't corelate it with my experiences. I own around 20 headphones and I bet one of these has a ringing issue. So my question is simple, could you give me an example of ringing, so I can hear what it is ?

    List of headhphones I can test :

    • HD700 (I mention it directly because it looks like takato14 said that it can have some)
    • HD598
    • ATH1000
    • ATH M50x
    • NAD VISO HP 50
    • Focal Spirit Profesionnal
    • Superlux 668B
    • Superlux 681
    • Ultrasone PRO2900

    Thanks
     
  4. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    HD700 for sure. Grab a Grado with bowls. A lot of the headphones you listed are actually good and free from any significant ringing.
     
  5. Bagged Milk

    Bagged Milk Friend

    Pyrate
    Joined:
    Oct 26, 2015
    Likes Received:
    100
    Trophy Points:
    43
    Location:
    Minnesota
    Are the Ultrasone Pro2900 included in the "good and free from any significant ringing." category? :confused:
     
  6. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Alright, thank you for your reply. It's a very good thing to know for the hd700 because it may explain why I tend to prefer the hd598 versus hd700. I'll try to catch it. Is there a clear way to identify ringing in a song ? I mean, how would you noticd this ringing while listening ?
     
  7. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    @Bagged Milk, I can't answer for the PRO2900 as I'm continuing my investigation. However, I started to analyze measurements by Tyll and some CSDs as I assume it is two possible ways to directly check whether a headphone has some or not.
    If I'm not mistaken, "ringing" can be listened when the diaphragm is still moving after a "long time". This "long time" is a reference to the "2ms" that @The Alchemist mentionned. So what I understood is that "ringing" is when you still hear a sound after a signal excited the diaphragm. This also explains why a square wave and CSD are good testing procedures to check wether the headphone has ringing effect or not.

    From here, I see three interesting points and I wish to discuss about them (plus to confirm my understanding of the phenomenum) :
    1. One can be more or less sensitive to ringing, depending on its capacity to perceive or by its natural "head filtering". It may explain some preferences although I think we are generally more tolerant to lows ringing but treble ringing is more seen as harsh, annoying and unpleasant (who said that a moscito may sound like a ringing ?)
    2. On a square wave response, if there are several peaks after the first overshoot, it may indicate ringing effect. Can you confirm ? Because I've been checking the measurement of the HD700 and it is like that. If it is the case, I am wondering about headphones presenting only two or three peaks after the overshoot. I think they have slight ringing effect (the Ultrasone PRO2900 may lay in this situation)
    3. Also I am wondering how much can we see ringing effect on a spectrum, because I think resonances may indicate some. Could you discuss this point too ?
     
  8. Bagged Milk

    Bagged Milk Friend

    Pyrate
    Joined:
    Oct 26, 2015
    Likes Received:
    100
    Trophy Points:
    43
    Location:
    Minnesota
    http://ko.goldenears.net/board/inde...&search_keyword=pro+2900&document_srl=1631011 Well... Goldenears has a CSD of the Pro2900. I was joking about how Purrin never said the Pro2900 when he was commenting on your cans.
     
  9. kapanak

    kapanak Canucklehead - Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    259
    Trophy Points:
    63
  10. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    @Bagged Milk thank you for your reply, also, I will just reformulate my question as simply as I can, because I may be too complicated :
    1. How do you see ringing effect on measurements ?
    2. How do you feel it while listening ?
     
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    2-3k
    Vocal shrillness.
    Guitar crunch.
    Shouty.

    4-5k
    Adds a bit of edge, bite, and excitement to the sound.
    Detail (macro / coarse).

    6k
    Generally tolerable, but unpleasant.
    Similar to 5k in that it adds some bite
    Hardness bordering on glare.

    7k
    Glare galore with some sibilance.
    Sharpness.

    8k
    Very bad. Nasty. Painful.
    Sibilance with some tizziness.

    9k-12k
    Adds definition.
    More tolerable, but still unpleasant if too much.
    Tizziness and sizzle.
     
  12. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Thank you very much Purrin, I really needed those to get closer to the phenomenom. Because I could always check a CSD and say : "oh yeah it looks like there's some ringing" but I just wanted to correlate those to practical experience. Again, thank you very much.
    You guys can add what you feel too, that would be awesome :).
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    FYI, I've revised the CSD visualizations slightly. Going back to the old Changstar presentation that I used years ago with some tweaks. Here are the changes:
    1. The max time has been reduced from 5ms to 3ms
    2. Floor lowered to -36db from -32db
    3. Incorporated a normalization process to account for peaks and overall volume of the frequency response to set the appropriate level from t=0.
    The CSD plots will look much less forgiving. I feel this is warranted because many of the CSD plots since V2 seemed to clean. Note that Ultrabike's max time is 2.5ms.
     
  14. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Okay good to know. I have some questions according to what you said @purrin.
    How do you set these parameters ?
    What are you willing to see ?
    If you have a dynamic headphones or an ortho, does make any difference ?
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Difficult to explain, but visualization / presentation of data is an art. Measurements are great and all, but useless without proper presentation.

    How I set these parameters is very much according to what I personally can hear, which does tend to be more sensitive than the average person. A few people have actually asked me to go back to the older presentations because they make the measurements look more "ugly".

    For example, lengthening the width of a FR or the time scale of a CSD can make the graph look "better". Extending the Y-axis amplitude can make deviations look "flatter". Manufacturers have been known to use these sorts of tricks and even more like excessive smoothing on FR/CSD or using linear instead of logarithmic scale for distortion (the db scale is logarithmic, not linear).

    Of course it is always possible to make the visualizations too revealing, where novices might take small deviations in the plots as something far more serious than they should be taken.

    It does not matter if the headphones are orthos or traditional dynamics. The presentation should be one that accounts for the worst measurement behaviors of both.
     
  16. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Thank you again @purrin this tells me a lot and especially because it is honest opinion.
    As you could see during this thread I am pretty new to this type of measurement and I am still trying to understand what is relevant or not. My feeling so far is that it looks like one of the best compromise you can find to see time and spectrum simultaneously.
    The more I dig into headphones and the more I feel the time dimension is underestimated, I think more is happening than what we can think.

    I have another question : I have the feeling that CSD is the metric to quantify/qualify headphones in the industry. Could you let me if it's true and, if it is, if there are specific reasons ? For instance, why not showing only the frequency response ? (I have some clues but I would be glad if you could answer without any suggestions, just honest opinion :) ).
     
  17. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Frequency response, CSD, and non-linear distortion are the trifecta of measurements for transducers. With enough experience listening at looking at the measurements, one can deduce a lot of what a transducer will sound like, or at the very least know that something is going to sound horrible to them.

    CSD plots are particularly useful for observing the behavior of a transducer from about 1kHz up. As discussed above, sometimes frequency response cannot capture a treble peak or ringing which manifests as a ridge in the CSD. CSD in addition to impulse response are excellent tools to observe how fast the driver dissipates energy.

    There has been some correlation between how resolving a transducer is (being able to reproduce low level information) and how low the CSD floor can go while looking clean between the 5kHz to 10kHz region. I'll dig up some examples later or just ask me again if I forget.
     
  18. Valolilol

    Valolilol Acquaintance

    Contributor
    Joined:
    Oct 29, 2015
    Likes Received:
    44
    Trophy Points:
    28
    Okay thank you very much @purrin I would be glad to see these correlations. I did an attempt to make some but I feel I needed a measurement setup, it's pretty hard to make it while using only others's measurements.
     
  19. keanex

    keanex Martian Bounty Hunter - Friend

    Pyrate
    Joined:
    Oct 1, 2015
    Likes Received:
    320
    Trophy Points:
    63
    O_O That looks scary as heck.

    As for measuring things and coming back with consistent and usable results, it's f'ing hard and I have the utmost respect for ultrabnike and purrin after trying my hand at it.
     
  20. Bagged Milk

    Bagged Milk Friend

    Pyrate
    Joined:
    Oct 26, 2015
    Likes Received:
    100
    Trophy Points:
    43
    Location:
    Minnesota
    all ultrasones measure scary as heck. :p
     

Share This Page