IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
Perception and Measurement of Headphone Sound Quality
Nessuno
post Apr 25 2013, 21:46
Post #26





Group: Members
Posts: 422
Joined: 16-December 10
From: Palermo
Member No.: 86562



QUOTE (extrabigmehdi @ Apr 25 2013, 13:58) *
A "neutral" frequency response, is the first thing that people ask when looking for a headphone.
But it's not enough to warrant good sound quality.

All the more, there's no general consensus that "neutral" means "flat". Think for example to different design choices like free field vs diffuse field equalization.


--------------------
... I live by long distance.
Go to the top of the page
+Quote Post
DVDdoug
post Apr 25 2013, 21:47
Post #27





Group: Members
Posts: 2568
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
The website innerfidelity provide different measurements for each headphones related to sound : frequency response, thd+ noise, impulse response, 30hz square wave response, 300 hz square wave response. So basically you are saying that I only need to look at the graph for freq response & THD + noise to get an idea of sound quality ?
If you see something "funny" in the square wave or impulse response that affects sound quality, it should aso show-up in the frequency response & distortion measurements.

In theory, I suppose you could have a driver that "rings" along with flat frequency response. But with a real-world mechanical resonance, there is normally increased amplitude at resonance.

As far as I know, Ethan Winer's list of 4 factors covers everything we can hear!

You can actually do things to mess-up a square wave without affecting the sound (such as run it through an all-pass filter to shift the relative phase at different frequencies without altering the spectral content).

This post has been edited by DVDdoug: Apr 25 2013, 21:54
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 25 2013, 22:42
Post #28





Group: Members
Posts: 408
Joined: 15-August 09
Member No.: 72330



QUOTE (Nessuno @ Apr 25 2013, 20:46) *
All the more, there's no general consensus that "neutral" means "flat". Think for example to different design choices like free field vs diffuse field equalization.

These considerations are a bit too technical for me. I guess that when you use the word "flat", you means that it sounds "flat" to the listeners,
and not a "flat" frequency response. I already before tried to "flatten" the freq response of a headphone, and the result is not great.
I 've never heard of free field equalization before, I think all modern headphones use diffuse field equalization.

This post has been edited by extrabigmehdi: Apr 25 2013, 22:47
Go to the top of the page
+Quote Post
greynol
post Apr 25 2013, 23:04
Post #29





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



If your pinna doesn't affect the sound when wearing headphones like is does when you're listening to the world around you without headphones, then flattening the response of the phones with an EQ will not provide the desired effect. I imagine it's also possible not to get the desired effect if the EQ adversely alters the phase. Though many claims about sound quality as it relates to phase response seem a bit dubious to me, so I say this only tepidly.


--------------------
I should publish a list of forum idiots.
Go to the top of the page
+Quote Post
xnor
post Apr 25 2013, 23:13
Post #30





Group: Developer
Posts: 598
Joined: 29-April 11
From: Austria
Member No.: 90198



From my experience phase is really a non-issue in single driver headphones. Only in multi driver head/earphones with crossovers you can see that the energy is not focused in the first peak in the impulse response.
Still, experiments with (multiple iirc) allpass filters shows that phase shifts are very hard/impossible to detect even with headphones as long as you listen to music and not some test signals.

This post has been edited by xnor: Apr 25 2013, 23:18
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 26 2013, 00:37
Post #31





Group: Members
Posts: 408
Joined: 15-August 09
Member No.: 72330



QUOTE (greynol @ Apr 25 2013, 22:04) *
If your pinna doesn't affect the sound when wearing headphones like is does when you're listening to the world around you without headphones, then flattening the response of the phones with an EQ will not provide the desired effect. I imagine it's also possible not to get the desired effect if the EQ adversely alters the phase.


Well my pinnas were remodeled because of cosmetic surgery. Logically my pinnas doesn't affect sound the same way as before, but I was not aware of a change in my hearing.

QUOTE (xnor @ Apr 25 2013, 22:13) *
Still, experiments with (multiple iirc) allpass filters shows that phase shifts are very hard/impossible to detect even with headphones as long as you listen to music and not some test signals.


Having played with different kind of eq (minmum phase & linear phase) , I'm skeptical when you say that the phase shift are undetectable.
Go to the top of the page
+Quote Post
DVDdoug
post Apr 26 2013, 01:02
Post #32





Group: Members
Posts: 2568
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
Having played with different kind of eq (minmum phase & linear phase) , I'm skeptical when you say that the phase shift are undetectable.
And I'm skeptical that it is detectable. wink.gif

To run an ABX test, you'd need an equalizer with "normal" setting and a "linear phase" setting. And, that switch/setting could not affect the EQ settings. I doubt that such an equalizer exists (at least in not hardware). It's not "fair" to compare two different equalizers. And of course to be accepted on this forum, the test has to be blind and otherwise in compliance with TOS #8.

I don't have a reference, but it's also my understanding that tests have been done with all-pass filters (filters that alter phase relationships without alterning frequency response).
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 26 2013, 01:39
Post #33





Group: Members
Posts: 408
Joined: 15-August 09
Member No.: 72330



QUOTE (DVDdoug @ Apr 26 2013, 00:02) *
To run an ABX test, you'd need an equalizer with "normal" setting and a "linear phase" setting. And, that switch/setting could not affect the EQ settings. I doubt that such an equalizer exists (at least in not hardware). It's not "fair" to compare two different equalizers. And of course to be accepted on this forum, the test has to be blind and otherwise in compliance with TOS #8.


There's the eq ddmf lp10, that allow to set the "amount" of phase separately for each frequency band. (i.e 0 for minimum phase, 1 for linear phase, and any values between).
Would you consider using such eq for comparison "fair" ?

This post has been edited by extrabigmehdi: Apr 26 2013, 01:55
Go to the top of the page
+Quote Post
xnor
post Apr 26 2013, 02:29
Post #34





Group: Developer
Posts: 598
Joined: 29-April 11
From: Austria
Member No.: 90198



For all intents and purposes single driver headphones are (mostly) minimum phase, so flattening the frequency response using a linear phase EQ with pre-ringing just doesn't make sense.
And I guess we all know that minimum phase filters are "perfectly reversible". So for equalization during playback linear phase is crap imho.

On audibility of phase:
http://www.audioholics.com/education/acous...dibility-part-2

edit:
QUOTE (extrabigmehdi @ Apr 26 2013, 00:37) *
Having played with different kind of eq (minmum phase & linear phase) , I'm skeptical when you say that the phase shift are undetectable.

What does a linear phase EQ have to do with hearing phase shifts? When you compare min. with linear phase EQs you are not just comparing phase shift but also the audibility of pre-ringing.

QUOTE (extrabigmehdi @ Apr 26 2013, 01:39) *
There's the eq ddmf lp10, that allow to set the "amount" of phase separately for each frequency band. (i.e 0 for minimum phase, 1 for linear phase, and any values between). Would you consider using such eq for comparison "fair" ?

Just use a min. phase EQ that has an all pass filter.

This post has been edited by xnor: Apr 26 2013, 02:34
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 26 2013, 12:30
Post #35





Group: Members
Posts: 408
Joined: 15-August 09
Member No.: 72330



QUOTE (xnor @ Apr 26 2013, 01:29) *
For all intents and purposes single driver headphones are (mostly) minimum phase, so flattening the frequency response using a linear phase EQ with pre-ringing just doesn't make sense.
And I guess we all know that minimum phase filters are "perfectly reversible". So for equalization during playback linear phase is crap imho.


Well, does a minimum phase eq allows to cancel the phase from the headphones ? If not, they may just add more phase, and using "linear phase" still makes sense.
And regarding the "pre-ringing", it can be minimized by using enough smooth eq curve which means , you don't try to "flatten" exactly the freq response.

But I don't try to "flatten" the freq curve anyways. According to headroom website:

QUOTE
A "natural sounding" headphone should be slightly higher in the bass (about 3 or 4 dB) between 40Hz and 500Hz. This compensates for the fact that headphones don't give you the physical punch or 'impact' that the sound waves from a room speaker have; so a slight compensation for increased bass response is needed for natural sound.

Headphones also need to be rolled-off in the highs to compensate for the drivers being so close to the ear; a gently sloping flat line from 1kHz to about 8-10dB down at 20kHz is about right. You'll notice all headphone measurements have a lot of jagged ups & downs (peaks & valleys) in the high frequencies; this is normal and mostly due to reflection cancellations in the folds and ridges in the outer part of the ear. Ideally however, the ups and downs of the frequency response should be fairly small and average out to a flat line. Large peaks or valleys over 3kHz in width usually indicate poor headphone response, and should be viewed as a coloring of the sound. Some small dips in the highs may actually be desirable and should exist in the 2kHz to 8kHz region.


http://www.headphone.com/learning-center/t...easurements.php


Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Apr 26 2013, 13:19
Post #36





Group: Members
Posts: 3701
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (DVDdoug @ Apr 25 2013, 13:17) *
QUOTE
In practical terms, i.e. when buying some phones for actual money; can we safely conclude that frequency graphs like the ones on Headroom are adequate indicators of sound quality?
Yes, unless there is audible distortion.

Ethan Winer identifies four things that affect/describe sound-reproduction quality:
Noise
Distortion
Frequency Response
Time Based Errors


Ethan's presentation of this matter lacks rigor and therefore encourages a certain amount of discussion that has limited benefits.

The three (3) things that affect sound reproduction quality are:

Linear distortion, which includes both frequency and phase response
Nonlinear distortion, which includes both amplitude distortion and time-domain distortion
Noises that are uncorrelated with the signal which includes random noise and deterministic noises.

QUOTE
Headphones don't generate any noise (other than mechanical noise when you move the cord, etc.),


The ultimate noise floor of headphones comes from the environment, which is the Brownian Motion of air molecules.

The noises you describe above are "Noises that are uncorrelated with the signal which includes random noise and deterministic noises". Mechanical noise is largely deterministic.

QUOTE
and they don't generate time-based errors either.


Except they do. Headphones with moving diaphragms are prone to generate some measurable amount of FM distortion which is nonlinear distortion and is time-based.

QUOTE
Distortion & frequency response are the only things left.


I submit that headphones like almost everything else in life generate some measurable amounts of the the three things I listed above. Now we can quibble whether my list has 3 elements or six elements (if we count the internal breakdown of each of the three points).

Behind the 3 points lies decade after decade of an development of an established technology (Audio) composed of principles, doctrine, tests and criteria. For example nonlinear distortion has the potential to add new frequencies, while linear distortion does not. Some flinch at the apparent oxymoron "Linear Distortion" but if you compare and contrast linear and nonlinear distortion, you grow wiser and more empowered.

This is all consistent with the publications and wisdom of the audio arts as handed down in various refereed publications over the decades. I think we do better trying to understand the wisdom of the giants whose shoulders we stand than to ignore them and their work.
Go to the top of the page
+Quote Post
markanini
post Apr 26 2013, 13:30
Post #37





Group: Members
Posts: 550
Joined: 22-December 03
From: Malmö, Sweden
Member No.: 10615



extrabigmehdi, youre not getting it. Just stop.
Go to the top of the page
+Quote Post
xnor
post Apr 26 2013, 13:46
Post #38





Group: Developer
Posts: 598
Joined: 29-April 11
From: Austria
Member No.: 90198



QUOTE (extrabigmehdi @ Apr 26 2013, 12:30) *
Well, does a minimum phase eq allows to cancel the phase from the headphones ? If not, they may just add more phase, and using "linear phase" still makes sense.

As I wrote, every minimum phase system has an inverse (that is causal and stable) so yes, for example, by correcting a dip at 500 Hz with a min. phase EQ you not only correct the magnitude but also the phase of the frequency response.

Think of minimum phase as seeing the lightning first and then hearing the thunder, and linear phase as hearing the thunder first, then seeing the lightning and finally hearing the thunder again. Linear phase still has its uses like antialiasing filters where the pre-ringing is inaudible or:


QUOTE
And regarding the "pre-ringing", it can be minimized by using enough smooth eq curve which means , you don't try to "flatten" exactly the freq response.

But I don't try to "flatten" the freq curve anyways. According to headroom website:

It's the same for minimum phase, hence the name minimum phase. Doing smooth adjustments obviously causes less phase shift, just like it causes little less linear phase shift/group delay/pre-ringing with linear phase filters.

The thing is that most headphones do not have such a smooth frequency response to begin with. For example, the HeadRoom graphs are heavily smoothed, resulting in the illusion that using smooth filters will fix everything.

But sure, I agree that even if you're using a limited graphic EQ you can improve most headphones.

This post has been edited by xnor: Apr 26 2013, 13:55
Go to the top of the page
+Quote Post
Nessuno
post Apr 26 2013, 14:05
Post #39





Group: Members
Posts: 422
Joined: 16-December 10
From: Palermo
Member No.: 86562



QUOTE (extrabigmehdi @ Apr 25 2013, 23:42) *
QUOTE (Nessuno @ Apr 25 2013, 20:46) *
All the more, there's no general consensus that "neutral" means "flat". Think for example to different design choices like free field vs diffuse field equalization.

I guess that when you use the word "flat", you means that it sounds "flat" to the listeners, and not a "flat" frequency response.

I meant exactly "flat frequency response", which is a precise concept and quite easy to quantitatively assess, whereas "neutral" is a qualitative concept and subject to opinions and interpretations that often differ from listener to listener. You were speaking of frequency response and used the word "neutrality" and what I wanted to say, in other words, was that a flat frequency response measurement, which for example is considered at least a good starting point to achieve subjective neutrality for speakers, is most often considered a wrong one for headphones (edit: should you ever encounter one... wink.gif).

But from other posts you seem to know already.

This post has been edited by Nessuno: Apr 26 2013, 14:16


--------------------
... I live by long distance.
Go to the top of the page
+Quote Post
extrabigmehdi
post Apr 26 2013, 15:58
Post #40





Group: Members
Posts: 408
Joined: 15-August 09
Member No.: 72330



QUOTE (xnor @ Apr 26 2013, 12:46) *
so yes, for example, by correcting a dip at 500 Hz with a min. phase EQ you not only correct the magnitude but also the phase of the frequency response.

This is great news, although it would be interesting if I could study further this point. There's the dsp "bbe d82 exciter" , that claims to "compensate" phase shift.
I guess that it's mostly a basic minimum phase eq then.

QUOTE
Think of minimum phase as seeing the lightning first and then hearing the thunder, and linear phase as hearing the thunder first, then seeing the lightning and finally hearing the thunder again.

Yes but if I re-use your analogy , the amount of "thunder" for "linear phase" is reparted before and after the "lightening" : i.e 50% before and after.
But from what I've gathered, we are more sensitive to "pre-ringing", than "post-ringing".

This post has been edited by extrabigmehdi: Apr 26 2013, 16:15
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Apr 26 2013, 16:26
Post #41





Group: Members
Posts: 3701
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (xnor @ Apr 26 2013, 08:46) *
But sure, I agree that even if you're using a limited graphic EQ you can improve most headphones.


That I agree with.

A lot of hand-wringing over phase is counter-productive because the ears are fairly insensitive to phase > 1 Khz as long as we are talking relatively small angles and applied to both channels identically.

The ear is also not all that sensitive to frequency response aberrations that only cover narrow bands:

http://home.provide.net/~djcarlst/abx_crit.htm
Go to the top of the page
+Quote Post
xnor
post Apr 26 2013, 19:03
Post #42





Group: Developer
Posts: 598
Joined: 29-April 11
From: Austria
Member No.: 90198



QUOTE (extrabigmehdi @ Apr 26 2013, 15:58) *
This is great news, although it would be interesting if I could study further this point. There's the dsp "bbe d82 exciter" , that claims to "compensate" phase shift.
I guess that it's mostly a basic minimum phase eq then.

This seems to process signals in a nonlinear fashion. Anyway it is definitely not a basic minimum phase EQ.

QUOTE
Yes but if I re-use your analogy , the amount of "thunder" for "linear phase" is reparted before and after the "lightening" : i.e 50% before and after.
But from what I've gathered, we are more sensitive to "pre-ringing", than "post-ringing".

I don't think this (50% before and after) is true. A quick check with a +6 dB peak at 4 kHz with Q=1.14 shows it's more like over 80% before and 80% after (for the samples to drop to about -80 dB).
But we're definitely more sensitive to pre-ringing because it's unnatural.

This post has been edited by xnor: Apr 26 2013, 23:41
Go to the top of the page
+Quote Post
markanini
post Apr 26 2013, 20:14
Post #43





Group: Members
Posts: 550
Joined: 22-December 03
From: Malmö, Sweden
Member No.: 10615



QUOTE (Nessuno @ Apr 26 2013, 15:05) *
I meant exactly "flat frequency response", which is a precise concept and quite easy to quantitatively assess, whereas "neutral" is a qualitative concept and subject to opinions and interpretations that often differ from listener to listener. You were speaking of frequency response and used the word "neutrality" and what I wanted to say, in other words, was that a flat frequency response measurement, which for example is considered at least a good starting point to achieve subjective neutrality for speakers, is most often considered a wrong one for headphones (edit: should you ever encounter one... wink.gif).

But from other posts you seem to know already.

Possible concepts to conflate don't end here.

This post has been edited by markanini: Apr 26 2013, 20:14
Go to the top of the page
+Quote Post
Martel
post Sep 12 2013, 08:24
Post #44





Group: Members
Posts: 553
Joined: 31-May 04
From: Czech Rep.
Member No.: 14430



QUOTE (DVDdoug @ Apr 25 2013, 19:17) *
Yes, unless there is audible distortion.

Ethan Winer identifies four things that affect/describe sound-reproduction quality:
Noise
Distortion
Frequency Response
Time Based Errors

Headphones don't generate any noise (other than mechanical noise when you move the cord, etc.), and they don't generate time-based errors either. Distortion & frequency response are the only things left.

What about stereo imaging of a headphone? How can you possibly judge that from a single FR chart? Even telling from two charts (one for each channel) is non-intuitive.


--------------------
IE4 Rockbox Clip+ AAC@192; HD 668B/HD 518 Xonar DX FB2k FLAC;
Go to the top of the page
+Quote Post
Nessuno
post Sep 12 2013, 09:55
Post #45





Group: Members
Posts: 422
Joined: 16-December 10
From: Palermo
Member No.: 86562



QUOTE (Martel @ Sep 12 2013, 09:24) *
QUOTE (DVDdoug @ Apr 25 2013, 19:17) *

Headphones don't generate any noise (other than mechanical noise when you move the cord, etc.), and they don't generate time-based errors either. Distortion & frequency response are the only things left.

What about stereo imaging of a headphone? How can you possibly judge that from a single FR chart? Even telling from two charts (one for each channel) is non-intuitive.

Frequency response, even when measured by a dummy head, is attributable only to the particular transducer and encasing used, so it characterizes objective properties, while stereo imaging depends strongly, amongst other factor, from the recording (i.e. conventional vs binaural) and the HRTF of the listener (and even expectations, experiences of live listening, mood etc... wink.gif).

Generally speaking I think stereo imaging/illusion is the least measurable or otherwise characterisable and predictable behaviour, both for headphone and speakers. It's a field where in the end only personal subjective evaluation matters.


--------------------
... I live by long distance.
Go to the top of the page
+Quote Post
markanini
post Sep 12 2013, 10:01
Post #46





Group: Members
Posts: 550
Joined: 22-December 03
From: Malmö, Sweden
Member No.: 10615



QUOTE (Martel @ Sep 12 2013, 09:24) *
QUOTE (DVDdoug @ Apr 25 2013, 19:17) *
Yes, unless there is audible distortion.

Ethan Winer identifies four things that affect/describe sound-reproduction quality:
Noise
Distortion
Frequency Response
Time Based Errors

Headphones don't generate any noise (other than mechanical noise when you move the cord, etc.), and they don't generate time-based errors either. Distortion & frequency response are the only things left.

What about stereo imaging of a headphone? How can you possibly judge that from a single FR chart? Even telling from two charts (one for each channel) is non-intuitive.


What stereo image? Use a crossfeed DSP. wink.gif
Go to the top of the page
+Quote Post
xnor
post Sep 12 2013, 10:54
Post #47





Group: Developer
Posts: 598
Joined: 29-April 11
From: Austria
Member No.: 90198



Imaging with stereo recordings in headphones is a complex topic (I use to say: "there is no sound stage") since there's a mismatch between what our hearing expects and what is being output by the headphone drivers.

Based on anecdotes, that open, full-size headphones have the best imaging I'd say:
- "bigger" headphones incorporate the pinna more (IEMs bypass everything but the last part of the ear canal)
- open headphones have more crosstalk (IEMs have high isolation and seal each ear canal)

Those things result in more natural sound. Additionally, headphones with boosted bass may sound more spacious since low frequency content in stereo recordings is also closer to what our hearing expect than treble.

This post has been edited by xnor: Sep 12 2013, 10:54
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 2nd September 2014 - 07:26