• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Distortion Listening Test

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
It's actually too subtle to decide is it correlated with signal magnitude or not. But it's almost for sure correlated with overall distortions level. So at high distortions it's seems easier to hear than on low. Maybe someone can listen to it and correct me.
Alright
 

outerspace

Active Member
Joined
Aug 18, 2020
Messages
111
Likes
76
I had the same result with BI version of the test.
MUSIC klippel BI 2.PNGMUSIC klippel BI.PNG

But L(x) is a much harder. I don't even sure that this kind of distortion even exist in L(x). I hardly reach -54 dB after several takes.
MUSIC klippel Lx 2.PNGMUSIC klippel Lx.PNG

I've read description of test and as I understand they use simulation and things that we hear are artificially added distortions. We don't listen real speakers but different models of speaker's distortions (simulations). I contacted them about this "hiss" - is this part of added distortions or some flaw of the test. Also I want to notice daftcombo found audible "noise" in "- mode" of test. But I tested "full range speaker mode". I think the fact that standard and BI version have this and L(x) don't is the evidence that this is a part of added distortions (part of speaker model) and not just accident flaw. Let's wait the answer from klippel.
 

1lifeliveit

Member
Joined
Aug 14, 2020
Messages
32
Likes
13
I did this test on my stereo system via bluetooth from my phone. Does this mean that my system is fairly transparent?
 

Attachments

  • Screenshot_20200910-191123.png
    Screenshot_20200910-191123.png
    211.5 KB · Views: 165

j_j

Major Contributor
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
2,267
Likes
4,759
Location
My kitchen or my listening room.
I did this test on my stereo system via bluetooth from my phone. Does this mean that my system is fairly transparent?

Bluetooth isn't exactly transparent at best. So I wouldn't consider that terribly accurate. Bluetooth can also exacerbate other errors.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
I had the same result with BI version of the test.
View attachment 82187View attachment 82188

But L(x) is a much harder. I don't even sure that this kind of distortion even exist in L(x). I hardly reach -54 dB after several takes.
View attachment 82185View attachment 82186

I've read description of test and as I understand they use simulation and things that we hear are artificially added distortions. We don't listen real speakers but different models of speaker's distortions (simulations). I contacted them about this "hiss" - is this part of added distortions or some flaw of the test. Also I want to notice daftcombo found audible "noise" in "- mode" of test. But I tested "full range speaker mode". I think the fact that standard and BI version have this and L(x) don't is the evidence that this is a part of added distortions (part of speaker model) and not just accident flaw. Let's wait the answer from klippel.
If you’re relying on the hiss then you’re not listening and responding to the actual distortion part of the test. Sorry, I believe that invalidates your results.
 
Last edited:

outerspace

Active Member
Joined
Aug 18, 2020
Messages
111
Likes
76
If you’re relying on the hiss then you’re not listening and responding to the actual distortion part of the test.
Distortions in this test are simulated by creating a mathematical model of real speaker. This "hiss" is part of simulated distortions. I would take your point if it's some artifact of real speaker recording, but it's not. It's actual part of the test and I even think it's correlated with signal (but it's too subtle so I'm not 100% sure). Also interesting thing is nobody mention this particular distortion in the thread. Seems like nobody noticed it. I do and it's "invalidates my results". Interesting logic. Sounds like I'm cheating the test or something.
what equipment do you use?
ES9023+O2 amp+dt990. But I can hear it even with my cheap equalized in-ears. I think you can hear it too, just carefully pay attention to first seconds of the track.
 
Last edited:

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago

stevenswall

Major Contributor
Forum Donor
Joined
Dec 10, 2019
Messages
1,366
Likes
1,075
Location
Orem, UT
Distortions in this test are simulated by creating a mathematical model of real speaker. This "hiss" is part of simulated distortions. I would take your point if it's some artifact of real speaker recording, but it's not. It's actual part of the test and I even think it's correlated with signal (but it's too subtle so I'm not 100% sure). Also interesting thing is nobody mention this particular distortion in the thread. Seems like nobody noticed it. I do and it's "invalidates my results". Interesting logic. Sounds like I'm cheating the test or something.

ES9023+O2 amp+dt990. But I can hear it even with my cheap equalized in-ears. I think you can hear it too, just carefully pay attention to first seconds of the track.

While noise floor is certainly distortion, at some point anyone could hear it if they can amplify it enough... More useful to see if someone notices without a "tell" but good job on identifying that.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
This "hiss" is part of simulated distortions. I would take your point if it's some artifact of real speaker recording, but it's not. It's actual part of the test.
I was unable to find a reference to noise or hiss in the link you provided. Can you point it out for us?
 

outerspace

Active Member
Joined
Aug 18, 2020
Messages
111
Likes
76
It's was different test mode and different part of track. So nobody mention this particular thing that I'm talking about.

I was unable to find a reference to noise or hiss in the link you provided. Can you point it out for us?

...At high amplitudes, real loudspeakers produce other kinds of distortion, which should also be investigated systematically. This can be done by using a nonlinear model of the loudspeaker that is able to synthesize loudspeaker output in the large signal domain...
...By measuring the parameters of the model, it is possible to perform a real-time simulation for any input, like music or artificial test signals, and to provide the distortion components separated from the ideal linear output...


This quote says all distortions are synthesized by nonlinear model of speaker.

So two factors can prove it's part of distortions. First, they use mathematical model to simulate distortions. Second, this thing is absent in in L(x). So I doubt this is unintended artifact, I think it's part of the test. But klippel didn't answer my email about the topic to be 100% sure.

Also I think there is may be terminological problem. Do you hear this distortion? How you can describe it?

While noise floor is certainly distortion, at some point anyone could hear it if they can amplify it enough...
Are you sure it's just a noise floor? Why it's absent in in L(x)? Audibility of many types of distortions can be improved by amplifying.
 
Last edited:

Francis Vaughan

Addicted to Fun and Learning
Forum Donor
Joined
Dec 6, 2018
Messages
933
Likes
4,697
Location
Adelaide Australia
It seems very unlikely that any noise is intentionally part of the distortion modelling. The modelling described in the Klippel paper is simply a matter of dynamically managing the transfer function and convolving the input signal with that function. They model all the usual suspects, suspension non-linearity, inductance non-linearity, thermal effects and so on. All of these just add terms to a continuous function. There is no stochastic noise.
However, the creation of the convolved output may well have not been done quite as cleanly as one might hope, and there exists the possibility that there are processing artefacts present. The obvious one would be dither. Or even the lack of dither. Either way, getting the dithering wrong could easily result in an unintentional audible tell.
 

oursmagenta

Active Member
Forum Donor
Joined
Jan 19, 2021
Messages
161
Likes
187
Location
France
Oh boy, I stumbled upon this thread, did the klippel test (Music J. Stone, full range -> -21 db), and now I feel almost ashamed of my hearing capability. Being average (median to be more precise) is the worst human condition right ?

Does it mean that now I'm "entitled" to buy crappy audio equipment, I mean -21 db is equivalent to almost 9 % of distortion?

I can't figure out the actual meaning of the test. The reference for the 0db is a "real" speaker with already some distortion with say a level of -Xdb.

So if I can tell the difference up to -21db (it seems that the -21db figure means that the "real" speaker distortion has been tone down by 21db), it means that I can tell the difference up to (-X - 21)db of distortion.

But if the -Xdb of distortion for the "real" speaker is -80db, which translate into a -101db audibility threshold, it is not the same as a -Xdb equal to -20db which translate into audibility of -41db of distortion.

So what is the actual distortion contained in a test file labeled as -21db ?

N.B: test done with a E30 -> L30 -> HD600
 
Last edited:

ZööZ

Addicted to Fun and Learning
Joined
Nov 10, 2020
Messages
553
Likes
352
Distortions in this test are simulated by creating a mathematical model of real speaker. This "hiss" is part of simulated distortions. I would take your point if it's some artifact of real speaker recording, but it's not. It's actual part of the test and I even think it's correlated with signal (but it's too subtle so I'm not 100% sure). Also interesting thing is nobody mention this particular distortion in the thread. Seems like nobody noticed it. I do and it's "invalidates my results". Interesting logic. Sounds like I'm cheating the test or something.

ES9023+O2 amp+dt990. But I can hear it even with my cheap equalized in-ears. I think you can hear it too, just carefully pay attention to first seconds of the track.

Is your dt990 equalized? I find that unequalized dt990 magnifies such a hiss
(even amir said he noticed such a thing https://www.audiosciencereview.com/...ro-review-headphone.19975/page-10#post-658325)
 

ZööZ

Addicted to Fun and Learning
Joined
Nov 10, 2020
Messages
553
Likes
352
Just took the test first time and after the first fail I realized that propably should have set the headphones louder (HIfiman HE4XX). Hindsight 20-20
Thought I was doing well until the first fail and somehow after that the test seemed harder
Screenshot_2021-02-08 Listening Test(3).pngScreenshot_2021-02-08 Listening Test(4).png

(Maybe I was over confident as I got lazy with the screencaps so just up until the second fail)

EDIT
Did the two tone test
Screenshot_2021-02-09 Listening Test(1).pngScreenshot_2021-02-09 Listening Test(2).png
 
Last edited:

outerspace

Active Member
Joined
Aug 18, 2020
Messages
111
Likes
76
Is your dt990 equalized? I find that unequalized dt990 magnifies such a hiss
(even amir said he noticed such a thing https://www.audiosciencereview.com/...ro-review-headphone.19975/page-10#post-658325)
Yes, they are equalized to flat according to oratory1990 measurements + individual peaks correction. And what is more, I can hear this distortion with my HD595 and all my in-ears, on normal volume. So it's not about equalization or high amplification. You just need high attention.
 
Last edited:

ZööZ

Addicted to Fun and Learning
Joined
Nov 10, 2020
Messages
553
Likes
352
I tried couple of the test with iems today (in a noisy environment though) and after a certain db treshold the hard thing wasn't to hear the distortion, but to tell which was more distorted. There couldn't be distortion in the control track also? (I'm quite positive I was only imagining that they both were somehow distorted)
 
Last edited:

Vaskis

Member
Joined
Jan 26, 2021
Messages
27
Likes
15
1612907416822.png

Basic setup pc+cheap headphones. Result -33. First time listener. Must admit those sounds not really pleasant to listen to.
1612907416822.png
 

Sharur

Senior Member
Joined
Apr 10, 2021
Messages
476
Likes
213
1619051890955.png

Topping E30 > NAD C316BEE V2 > DT880 600 Ω
 
Last edited:
Top Bottom