• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Binaural blind comparison test of 4 loudspeakers - II

Which speaker comes closest to the original recording?

  • Speaker A

    Votes: 7 25.9%
  • Speaker B

    Votes: 4 14.8%
  • Speaker C

    Votes: 3 11.1%
  • Speaker D

    Votes: 13 48.1%

  • Total voters
    27
  • Poll closed .

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
Thomas A wrote:
"As Toole and others already have said several times:
A: A speaker with a flat on-axis and smooth off-axis is the most preferred, when evaluated in mono with music.
B: The same speaker sound different when in stereo setup playing the same music in mono, because of the stereo system errors (comb filtering).

So if A is true, the same preference for B cannot at the same time be true. You need another frequency response to mimic the mono speaker setup.

Tooles solution is to add a center channel."


Isn't it a bit strange to use monophonic material to evaluate a stereo pair of speakers dedicated to stereophonic music material?

One could argue so. However very much of music content lies in the center of the phantom image, such as voices. There is just no optimal solution other than to add a center channel if you want correct timbre across the whole image. For stereo playback, there are only compromises which questions the optimal response of a +/- 0.0 dB variation of the on-axis response.
 

Tangband

Addicted to Fun and Learning
Joined
Sep 3, 2019
Messages
599
Likes
626
If you put a speaker with a linear response +/-1.5 dB against speakers with +/- 5 dB it will be quite certain which will win.

With respect to stereo system corrections you need to test speakers that are within +/- 1.5 dB. Such a test has not been performed in the Scientific literature. Yet.
Dont you wonder why it havenˋt been done yet ? Who are you refering to, and why do you know that this statement is even true ?
The loudspeaker I think you are refering to is a very good one, but it has not a good frequency respons off axis, meaning the directivity is gonna be worse than GRIMM LS1, Genelecs and Revels.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
Dont you wonder why it havenˋt been done yet ?
The speaker I think you are refering to is good, but it has not a good frequency respons off axis, meaning the directivity is worse than GRIMM LS1, Genelecs and Revels.
I am referring to the Revels here.

The experiment that needs to be done is to have three speakers with as good response as possible placed in center and stereo setup, say +/- 0.5 dB on axis and likewise smooth off axis

Play pink noise alternating the center and stereo. Adjust the stereo speakers to the same timbre as the center speaker using EQ.

Average the curve of the linear and EQed one.

Go for blind tests using various speakers.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
Dont you wonder why it havenˋt been done yet ? Who are you refering to, and why do you know that this statement is even true ?
The loudspeaker I think you are refering to is a very good one, but it has not a good frequency respons off axis, meaning the directivity is gonna be worse than GRIMM LS1, Genelecs and Revels.

It is old knowlegde, also in Tooles book. Stereo dilemma.
 

Tangband

Addicted to Fun and Learning
Joined
Sep 3, 2019
Messages
599
Likes
626
It is old knowlegde, also in Tooles book. Stereo dilemma.
Here is the source behind the need for stereo system compensation. The Guru qm60. As you can see- rather good measurement result but not a star at good directivity. You dont se a trace of the claimed needed peak at 1,4 kHz , but you can see a peak off axis between 3-5 kHz at 60 degree. The constructor have apparently compensated that peak off axis with a small dip on axis between 3-5 kHz.

GRIMM gets away with the dilemma of combining an 8 inch woofer with a small tweeter using a waveguide and crossing at 1,5 kHz , with is low but its needed if the result gonna be good directivity.

Regarding Guru qm60: If the directivity is really good, you dont need to compensate the on axis response with the off axis, like they did .:)

09360319-E4C2-4AD5-B8B9-D846812927F4.png
 
Last edited:

Juhazi

Major Contributor
Joined
Sep 15, 2018
Messages
1,554
Likes
2,496
Location
Finland
Creating L/C/R and using some eg. Dolby ProLogic II algorithm is not the right way to test hifi stereo loudspeakers. Aren't they suppposed to work best as stereo pair with stereophonic source? Voicing (if done) most likely is aimed for that scenario. And in test setup every subject (pair) is set up similarly in the room. Monitor speakers might be used as single, stereo or multichannel, and are sold as single units.

Harman studies were done with single speaker, I understand mainly to minimize room effect. Toole has said, that single speaker assessment correlates with stereo preference. That's fine for me. But still, stereophony makes more random variables to the game, and personal preferences of listeners regarding stereo imaging are huge. Most magazine reviews are done with a stereo pair, which explains some strange reports. And if there are measurements, they are for a single speaker.

99,9% of popular music is mixed and mastered in stereo. They use multichannel practically only for some classical or electronic extravaganza.
 

oivavoi

Major Contributor
Forum Donor
Joined
Jan 12, 2017
Messages
1,646
Likes
1,629
Location
Oslo, Norway
It is old knowlegde, also in Tooles book. Stereo dilemma.

It's of course true that the stereo image appears timbrally different from the direct sound of the loudspeakers, due to comb filtering. What does not follow from this is that loudspeakers will sound better if they attempt to "correct" for this though global adjustments of the direct sound from the loudspeakers. I don't think there have been any tests which show that speakers which deviate from flat in certain areas are preferred? Here's an interesting (and long!) thread about an attempt to use DSP to equalize/brighten ONLY the parts of the signal which is summing between the loudspeakers, and not the rest - which makes much more sense in my view: https://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center.html

Furthermore, room reflections do a whole lot to mitigate this problem, again according to Toole (which aligns with my experience). The phantom/stereo image becomes really dull if one mostly hears the direct sound, either due to very narrow directivity or due to heavy room treatments. With loudspeakers with wide and even directivity this is much less of a problem.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
Here is the source behind the need for stereo system compensation. As you can see- rather good measurement result but not a star at good directivity. You dont se a trace of the claimed needed peak at 1,4 kHz , but you can see a peak of axis between 3-5 kHz at 60 degree. The constructor have apparently compensated that peak off axis with a small dip
on axis between 3-5 kHz.

If the directivity is really good, you dont need to compensate like that.:)

View attachment 160459

And I am quite confident that a QM60 or pI60s will come up as a top speaker also in a binaural contest. As you know 2 kHz dip which is had at +/-23° stereo setup cannot be compensated. You can only compensate partly by not exaggerating the response 3-4 kHz and 7-8 kHz.

Not one speaker is equal, and off-axis dips can be found even in Genelecs.

 
Last edited:

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
It's of course true that the stereo image appears timbrally different from the direct sound of the loudspeakers, due to comb filtering. What does not follow from this is that loudspeakers will sound better if they attempt to "correct" for this though global adjustments of the direct sound from the loudspeakers. I don't think there have been any tests which show that speakers which deviate from flat in certain areas are preferred? Here's an interesting (and long!) thread about an attempt to use DSP to equalize/brighten ONLY the parts of the signal which is summing between the loudspeakers, and not the rest - which makes much more sense in my view: https://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center.html

Furthermore, room reflections do a whole lot to mitigate this problem, again according to Toole (which aligns with my experience). The phantom/stereo image becomes really dull if one mostly hears the direct sound, either due to very narrow directivity or due to heavy room treatments. With loudspeakers with wide and even directivity this is much less of a problem.
Yes reflections do mitigate the stereo problem. This is also why any corrections would be within the spec of most speakers.

And as I mentioned, no such tests have been done. We assume that a perfectly linear response from a mono speaker also is the best speaker in stereo. But this assumption has not been challenged in any greater detail.
 

oivavoi

Major Contributor
Forum Donor
Joined
Jan 12, 2017
Messages
1,646
Likes
1,629
Location
Oslo, Norway
And as I mentioned, no such tests have been done. We assume that a perfectly linear response from a mono speaker also is the best speaker in stereo. But this assumption has not been challenged in any greater detail.

Agreed. It would be really interesting if somebody could make a binaural recording of a pi60, for example, and compare it to a SOTA speaker with flat response, such as the Grimm LS1. OR do a binaural recording of the LS1 where it is flat, and another where a so-called compensation for comb filtering is equalized in.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
Agreed. It would be really interesting if somebody could make a binaural recording of a pi60, for example, and compare it to a SOTA speaker with flat response, such as the Grimm LS1. OR do a binaural recording of the LS1 where it is flat, and another where a so-called compensation for comb filtering is equalized in.

It will be a hard test binaurally but certainly possible. I have suggested previously to test the same speakers with different EQ. Both within +/- 1.5 dB. One where there is a peaking at 3-4 kHz and 7-8 kHz, and one with the opposite EQ, i.e. dips at 3-4 khz and 7-8 kHz.

I could perhaps do it but I would need to a dummy head for optimal results.
 

Juhazi

Major Contributor
Joined
Sep 15, 2018
Messages
1,554
Likes
2,496
Location
Finland
We humans are adapted to comb filtering of stereo source, or better to sound perception with two ears in general. Time and phase difference are used to sound localization (typically of a single source), which is naturally challenged by reflections in the environment.

Baking in comb filtering compensation in a speaker's response is mission impossible. It happens at such short wavelengths that individual variance of head dimensions and earlobe (pinna) shape will harm that. If smoothing is used, it will change tonality too much. Speaker setup - distance between them and to listener varies even more.

Some modern authorities:

Perhaps someone else with me remembers, when stereo sets started to become popular at homes, in 1960s -early 70s? Typically speakers were positioned at sofa's ends and visitor was asked to sit in the middle of sofa. Then stereo from LPs sounded really strange - or wonderful!
 
Last edited:

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
It will be
We human are adapted to comb filtering of stero source, or better to sound perception with two ears in general. Time and phase difference are used to sound localization (typically of a single source), which is naturally challenged by reflections in the environment.

Baking in comb filtering compensation in a speaker's response is mission impossible. It happens at such short wavelengths that individual variance of head dimensions and earlobe (pinna) shape will harm that. If smoothing is used, it will change tonality too much.

Some modern authorities:

Perhaps someone else but me remembers, when stereo sets started to became popular at homes? Typically speakers were positioned at sofa's ends and visitor was asked to sit in the middle of sofa. Then stero from LPs sounded really strange - or wonderful!
It will be a judgement of what is "linear" or not. One could call my speakers linear on-axis with +/- 2.5 dB, but could also be called "stereo compensated". My years of listening and DIY have at least made me satisfied.

my speakers on axis 50 cm.png
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
I tried with my simple setup at home (again) using OM1/Omni1 microphones taped to my ears. At LP where the steady-state curve is sloping a la Harman. Microphones are not calibrated to diffuse field which may cause loss in the HF region.

The first file is the raw file and must be listened to headphones that have no bass elevation or treble drop a la Harman.


The second file is the raw file calibrated to be used with Harman target curve headphones.


Clearly there are much more of the "room" heard in my recordings.
 

Nicolaas

Active Member
Joined
Feb 2, 2020
Messages
116
Likes
96
Just to be clear, this isn't meant as a derogatory comment. I myself have been wearing in-ear microphones under a couple of headphones today.
Just a reflection on what a... peculiar bunch of people we are. :)
I like the scientific experiment too. But picturing the experiment together with your comment was very funny to me...
 

PierreV

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
1,257
Likes
3,852
Headphones... oh my!
Just for fun, I tried two open back headphones on top of the roland binaural microphones (Sennheiser HD600 vs Audio Technica ATH-ADG1X).
That's the top chart. Bottom chart is the previously posted result with two different set of speakers. That's not on the same piece of music - although I guess I should try that as well...

1634842298928.png


Those microphones are quite OK it seems. If one compares the recording I made with Amir's results, the HD600 is very easy to identify.

1634842976488.png
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
1,323
Likes
868
Location
Sweden
BTW, I made a simple recording of pink noise: a mono center speaker vs stereo speakers, only "left ear" below in mono.


Far from ideal experiment, but can you hear the timbral shift?
 
Top Bottom