or Connect
Head-Fi.org › Forums › Equipment Forums › Sound Science › Distortion measurements on headphones
New Posts  All Forums:Forum Nav:

Distortion measurements on headphones

post #1 of 16
Thread Starter 

Hi to Everyone !

 

I see a lot of frequency response graphs in the web on HPs

But no distortion measurements

I am currently looking for a HP set up with a very low distortion expecially in the midrange

My feeling is that a very low distortion in the midrange gives the best intellegibility of words in a speech, my very first need

Actually I have to listen for prolonged times ... a sort of full immersion in english plays with a lot of dialogues (about 2 hours each)

Not only, If the sound gets distorted I think it could become very fatiguing in the long term

In conclusion, which are HPs particularly noted for their very low distortion ?

Of course if anyone knows of distortion measurements I would be very interested

 

Thank you a lot and kind regards,

 

gino

post #2 of 16

Here, knock yourself out:

http://www.innerfidelity.com/headphone-data-sheet-downloads

post #3 of 16
Quote:
Originally Posted by DaBomb77766 View Post

Here, knock yourself out:

http://www.innerfidelity.com/headphone-data-sheet-downloads


Yes, these do include distortion graphs (vs. frequency), measured at both 90 and 100 dB SPL. Note however that the graphs show THD+N, rather than just THD, so in the mid to higher frequency range (where the distortion is usually the lowest) the measurements are often dominated by the noise of the test equipment.

 

post #4 of 16
Thread Starter 

 

Thank you so much indeed !

It is realy what I was looking for

By the way, do you agree on the fact the intellegibility of speech is related to a low distortion in the voice frequency range ?

Just to know what to look at

Thanks again and kind regards,

 

gino

post #5 of 16

If you mainly care about the clarity of speech, then the linearity of the frequency response in the midrange to lower treble (maybe a few hundred Hz to about 5 kHz) is also important.

 

post #6 of 16
Thread Starter 
Quote:
Originally Posted by stv014 View Post


Yes, these do include distortion graphs (vs. frequency), measured at both 90 and 100 dB SPL. Note however that the graphs show THD+N, rather than just THD, so in the mid to higher frequency range (where the distortion is usually the lowest) the measurements are often dominated by the noise of the test equipment.

 


Excuse me, you say

in the mid to higher frequency range (where the distortion is usually the lowest) the measurements are often dominated by the noise of the test equipment.

So do you mean that the actual distortion of most of the HPs in the mid to high range is lower than in the graph ?

Anyway impressive are the figures for the HE-60 ... even better, if I am not wrong, of the mighty HD800 !

Thanks again and kind regards,

 

gino

 

 

 

 

 

 

 

post #7 of 16
Quote:
Originally Posted by ginetto61 View Post

 

Excuse me, you say

in the mid to higher frequency range (where the distortion is usually the lowest) the measurements are often dominated by the noise of the test equipment.

So do you mean that the actual distortion of most of the HPs in the mid to high range is lower than in the graph ?

 

Yes, when you see "noisy" lines on the graph and the THD+N at 100 dB being lower than at 90 dB, then the actual distortion is often better than what is shown. Although if it is low enough to be outweighed by the noise, then it is most likely good enough for speech anyway.

 

post #8 of 16
Thread Starter 

 

Thank you very much again !

 

I understand that most of the evaluations on sound quality of HPs are usually carried out by listening, if not all

My feeling is that distortion measurements of this kind could be more objective/reliable

I do not understand why they are not very popular

Are they so difficult to perform ?

A driver that distorts actually adds frequencies not present in the test signal

From what I see I think that the Sennhheiser HE-60 must be a really exceptional headphone, and not only in terms of distortion but also sound quality

 

Kind regards,

 

gino

 

 

 


Edited by ginetto61 - 1/29/12 at 8:26am
post #9 of 16

Honestly, any higher-end monitoring headphones should suffice.  It really depends on how much you're looking at spending.  Reproducing spoken voice isn't terribly difficult for most competent headphones.

post #10 of 16

Also see here:

http://www.geocities.jp/ryumatsuba/review.html

 

It's in Japanese but the graphs are graphs.  Also I don't know how good the test setup is, as some results are a bit wonky.

 

HD650-Phase.gif

 

 

The graph at the bottom is the second harmonic (blue) and third harmonic (green) distortion over frequency...at least at some output level.  Note that the values are plotted +40 dB referenced to the values on the left axis.

post #11 of 16
Thread Starter 
Quote:
Originally Posted by DaBomb77766 View Post

Honestly, any higher-end monitoring headphones should suffice.  It really depends on how much you're looking at spending.  

Reproducing spoken voice isn't terribly difficult for most competent headphones.

 

Excuse me, do you deem for instance a AKG K242HD a higher-end monitoring HP ?   because it is the one I am using now

Buy the way I have also a JVC 900, an old Sennheiser 565 Ovation, an old AKG K280 Parabolic

I have also a Grado sr200 but unfortunately a driver got damaged

The 242 seems to me the better one overall

 

Thank you very much

Kind regards,

 

gino

 

post #12 of 16
Thread Starter 
Quote:
Originally Posted by mikeaj View Post

Also see here:

http://www.geocities.jp/ryumatsuba/review.html

 

It's in Japanese but the graphs are graphs.  Also I don't know how good the test setup is, as some results are a bit wonky.

 

HD650-Phase.gif

 

 

The graph at the bottom is the second harmonic (blue) and third harmonic (green) distortion over frequency...at least at some output level.  Note that the values are plotted +40 dB referenced to the values on the left axis.


Thank you very much indeed !  this is what I was lloking for really !

At least I am not the only one obsessed with distortion figures

I will go through all the graph.  It is too much interesting.

I see that the HD650 is splendid, as it is the HD595

Actually I am thinking to get a HD555.  It should have the same driver of the 595 and, hopefully, similar distortion figuresrolleyes.gif at a lower price biggrin.gif

 

Thank you so much again

Kind regards,

gino

 

P.S.  my reasoning being:

1)  low distortion is a must for good sound quality

2)  the HD595 has a very low distortion

3)  the HD555 has the same driver of the HD595

4)  the HD555 should have similar distortion figures compared to the HD595, expecially in the range of the voice (i.e. 200-4000Hz)

 

 

 

 

 

 


Edited by ginetto61 - 1/30/12 at 9:23am
post #13 of 16
Thread Starter 
Quote:
Originally Posted by mikeaj View Post
 

Also see here:

 

http://www.geocities.jp/ryumatsuba/review.html

 

It's in Japanese but the graphs are graphs.  Also I don't know how good the test setup is, as some results are a bit wonky.

 

HD650-Phase.gif

 

 

The graph at the bottom is the second harmonic (blue) and third harmonic (green) distortion over frequency...at least at some output level.  Note that the values are plotted +40 dB referenced to the values on the left axis.

 

 

Hi just to say that i think that this link of an extraordinary interested.

Unfortunately i have to open all the links and see the measurements, but i am very very willing to do so

In the weekend i will do because i have already seen something of astonishing

as the extremely low distortion of the V-MODA headphones ... amazing !!!!!

I am more and more convinced that very high quality sound implies extremely low distortion

I am sincerely excited

Thanks again indeed Mr. Mike !

Kindest regards,  gino


Edited by ginetto61 - 4/23/15 at 3:41am
post #14 of 16

Distortion is the best and only reliable way to test headphones. Period.

Comparing 2D visualizations is so primitive and inconclusive it is no wonder that manufacturers continue to sell magical wires and tubes. Comparison of devices is a computational problem in the same ways that banks calculate bond values and insurance companies calculate risk. Humans can't come close to doing a decent job - so I can't understand why so many insist on doing it. It is time for more audiophiles to embrace 21st century digital testing methodologies.  

 

Distortion is the central measurement 

 

Distortion or variation in relation to time and amplitude is the only measure that matters.  

 

There are just two basic measurement steps:
 

  1. Digitally sample a source signal S and output from a measured headphone M at a sufficiently high clock rate to reliably detect variances in both the amplitude axis and time axis.
     
  2. The resulting distortion is represented as a surface D that is the dimensional difference between S and M with respect to amplitude and time.    

 

All further investigation and analysis is performed on distortion surface D.

 

Various types of regression analysis of the distortion surface D create a distortion coefficient used to compare devices.  Algorithms that penalize time axis errors but forgive amplitude errors (a.k.a., frequency response variations) will be needed to properly flag 'accurate' versus 'colored' devices. This type of supervised machine learning is commonplace in fields as diverse as medicine and robotics. In these industries, errors rates can be 30% or higher.  Audio measurement - where % error rates are in the single digits - is relative child's play.

 

Human listening can be employed in a restricted manner. Descriptions of audiophile terms like "warm" and "forward" can be collected via surveys. The results can be used to categorize classes of distortion surfaces. There will be large variations in the results.  But humans have to do something, so you can at least ask them their opinion. Just don't use humans as measurement sensors.

You can also graph surface D as a visualization for human consumption if you wish. These images are inconclusive but large variations can at least be used to help filter really good or bad performers.  

 

 

We could do this type of testing now.   Not sure why we don't.


Edited by Gr8Desire - 4/23/15 at 8:30am
post #15 of 16
Thread Starter 
Quote:
Originally Posted by Gr8Desire View Post
 

Distortion is the best and only reliable way to test headphones. Period.

Comparing 2D visualizations is so primitive and inconclusive it is no wonder that manufacturers continue to sell magical wires and tubes. Comparison of devices is a computational problem in the same ways that banks calculate bond values and insurance companies calculate risk. Humans can't come close to doing a decent job - so I can't understand why so many insist on doing it. It is time for more audiophiles to embrace 21st century digital testing methodologies.  

 

Distortion is the central measurement 

 

Distortion or variation in relation to time and amplitude is the only measure that matters.  

 

There are just two basic measurement steps:
 

  1. Digitally sample a source signal S and output from a measured headphone M at a sufficiently high clock rate to reliably detect variances in both the amplitude axis and time axis.
     
  2. The resulting distortion is represented as a surface D that is the dimensional difference between S and M with respect to amplitude and time.    

 

All further investigation and analysis is performed on distortion surface D.

 

Various types of regression analysis of the distortion surface D create a distortion coefficient used to compare devices.  Algorithms that penalize time axis errors but forgive amplitude errors (a.k.a., frequency response variations) will be needed to properly flag 'accurate' versus 'colored' devices. This type of supervised machine learning is commonplace in fields as diverse as medicine and robotics. In these industries, errors rates can be 30% or higher.  Audio measurement - where % error rates are in the single digits - is relative child's play.

 

Human listening can be employed in a restricted manner. Descriptions of audiophile terms like "warm" and "forward" can be collected via surveys. The results can be used to categorize classes of distortion surfaces. There will be large variations in the results.  But humans have to do something, so you can at least ask them their opinion. Just don't use humans as measurement sensors.

You can also graph surface D as a visualization for human consumption if you wish. These images are inconclusive but large variations can at least be used to help filter really good or bad performers.  

 

 

We could do this type of testing now.   Not sure why we don't.

 

Hi and thank you very much indeed for the very valuable advice

I am on the same boat.  

I would like to see much more distortion measurements on headphones.  And i think that the methodology  that you propose is extremely valid. 

Regarding your final question " why we/they don't" well, if we agree that the lowest the distortion the best the sound what could happen if a cheap headphone had extremely low distortion like another one 20 times its price ?

I will check the site to see the best ones that they have tested.

This for instance is not cheap at all but ... OMG !  i want it !

 

 

 

If you have any particular recommendation please i am all ears

Thanks again and kindest regards,  gino 

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Distortion measurements on headphones