What’s Your Triangle in Stereo Speaker Listening?

[LIΟN]

The thing to compare is the resulting joint HRTF from both left and right to one ear (either one) to the actual zero degree hrtf.

It's an interesting indictment for the phantom center.

Thank you for your quick reply. Please understand that I am very overwhelmed to keep up with your knowledge.
Can you make it easier for me?
I don't know if I understand right.
Is it correct that we should compare L+R sum response at a specific angle (30 degrees, 60 degrees...etc) and HRTF at a 0 degree angle (front mono center)?

 

[STC]

For this, actual ITD's are important, because ITD isn't actually the same at every frequency for a given angle.

You mean speed of sound is frequency dependent? I only thought it was ILD that based due to HRTF between the ears.

 

[j_j]

You mean speed of sound is frequency dependent? I only thought it was ILD that based due to HRTF between the ears.

While speed of sound varies SLIGHTLY, the irregular shape of the head, etc, causes delay to change a bit at different frequencies. The primary (mean) number is certainly due to the path around the head from the source direction, of course.

But it's not the speed of sound that matters much here, it's the irregular shape of the head and the path from one side to the other.

Lion, you have it, compare L+R at your chosen angle vs. zero degree mono. That will show you a few interesting differences. There is also a nonlinear time-domain effect in the ear but that is, I think, somewhat secondary to the path.

 

[LIΟN]

I compared it like @j_j said.
First of all, I apologize in advance for having to continue the mouse scrolling because there are a few images.
Before checking the exact ms, I will briefly check the phase. And I compare both 80cm / 300cm just in case.
The reference point is hrtf of 0 degree mono, and the other two angles (30 degrees, 60 degrees) are the criteria for the sum response.

This time, we will check the process without the change in frequency amplitude-minphase.

Now let's check for a specific frequency.
Based on high frequency, 30 and 60 degrees had itd(dealy) of 229.2us and 458.3us, respectively.

When i compare 50 Hz at 80cm 0 and 30° angles, it comes out to about 90us.
and 500hz has 80us?

When i compare 50 Hz at 80cm 0 and 60° angles, it comes out to about 160us.
and 500hz has 110us?

When i compare 50 Hz at 300cm 0 and 30° angles, it comes out to about 80us.
and 500hz hase 70us?

When i compare 50 Hz at 300cm 0 and 60° angles, it comes out to about 150us.
and 500hz has 130us?


Once again, I apologize for the many images

So, To sum up the numbers

(Compared to HRTF at zero degrees)

80cm 30degree : 50hz-90us 500hz-80us
80cm 60degree : 50hz-160us 500hz-110us
300cm 30degㄱee : 50hz-80us 50hz-70us
300cm 60degree : 50hz-150us 500 hz-130us

(I think we can ignore 500hz of information. I think it's difficult to compare it relative to 0 degree HRTF due to ILD's intervention)

 

[STC]

When i compare 50 Hz at 300cm 0 and 30° angles, it comes out to about 80us.
and 500hz hase 70us?

not sure why and how this is relevant as 50hz is practically impossible to localize so ITD and ILD is irrelevant but out of curiosity why a 50hz wave which is 6.88m wavelength should reach much later than 500hz?

@j_j do you have any reference for difference in ITD for different frequencies.

@Lion, please measure a single mono sound with both mics from the same spot or XY configuration to see if there is a delay in the capture of two channels. digital audio is serial so the could be some inherent delay before the second channel start to capture and also whether it takes longer for the microphones to register lower frequencies? Also could you share all ITD till 20kHz?

 

[STC]

80cm 30degree : 50hz-90us 500hz-80us
80cm 60degree : 50hz-160us 500hz-110us
300cm 30degrteee : 50hz-80us 50hz-70us
300cm 60degree : 50hz-150us 500 hz-130us

hopefully some experts can explain how and why this is so.

 

[LIΟN]

@Lion, please measure a single mono sound with both mics from the same spot or XY configuration to see if there is a delay in the capture of two channels. digital audio is serial so the could be some inherent delay before the second channel start to capture and also whether it takes longer for the microphones to register lower frequencies? Also could you share all ITD till 20kHz?

I set my starting impulse to be similar as much as I could (the slight difference would be between 2 and 8us, and of course, 30 and 60 degrees ITD were maintained).
I told you to do this, but I don't know why this results appeared.
It may also be somewhat meaningless to check the ITD of all frequencies against 0 degrees (although it may be possible to compare them by frequency in each response).
Because there will be a clear difference with ILD's intervention starting from approximately 150-200 Hz, so it seems difficult to estimate ITD accurately with 0 degree reference. (So actually, I think I can ignore the 500 Hz value in the body I just wrote.)

And in quite a few graphs, I'm stuck in chaos like I'm in a swamp, so maybe there's something I've made a mistake about.

 

[STC]

so it seems difficult to estimate ITD accurately with 0 degree reference.

I was just curious if there is inherent delay between left and right channel. As far I know, audio digital transfer is serial so there could be some inherent delay. Nevermind that’s not important. Let’s wait for JJ to give some citation. This will be very useful for better XTC.

 

[LIΟN]

I also had a little bit of a question, so I wanted to remove the variable more clearly.

This is the room reaction of the dummy head above.
And I made this completely linear. And apply the correction filter as it is

And I compared this to one ear at 0 and one at 30 degrees.
The response will be almost really linear now, with one earring a 0.03 difference and this may be what REW estimates as a really subtle difference in bass response.

Because it's like this even if there's no delay.
If the phase is off, it will look like below.

And clearly, we see that the combined impulse of both ears at 30 degrees and the impulse of 0 degrees start almost exactly the same.

???
There is a difference when compared to the sum of both ears.
I don't know anymore.

@STC @j_j
I've done what I can. Now I'm waiting for your knowledge

 

[STC]

I also had a little bit of a question, so I wanted to remove the variable more clearly.

View attachment 393261

This is the room reaction of the dummy head above.
And I made this completely linear. And apply the correction filter as it is

View attachment 393262




View attachment 393263

And I compared this to one ear at 0 and one at 30 degrees.
The response will be almost really linear now, with one earring a 0.03 difference and this may be what REW estimates as a really subtle difference in bass response.

View attachment 393264
Because it's like this even if there's no delay.
If the phase is off, it will look like below.
View attachment 393265


View attachment 393266

And clearly, we see that the combined impulse of both ears at 30 degrees and the impulse of 0 degrees start almost exactly the same.



View attachment 393267

???
There is a difference when compared to the sum of both ears.
I don't know anymore.

@STC @j_j
I've done what I can. Now I'm waiting for your knowledge

In ear measurement is difficult. Anyway, this topic is new to me and I will wait for the relevant citation form @j_j . Otherwise, I think I have gone off topic long ago.

Edit- Going through some of the papers I wonder if this is about ITD fluctuation mentioned by Mason and Ramsey.

 

[tmuikku]

Here is something interesting:

Proportion of effects by head-related transfer function and receiver position variation to interaural cross-correlation values

Interaural cross-correlation coefficient (IACC) is an objective room-acoustic parameter associated with the spatial impression and subjective quality of the aco

pubs.aip.org

Study says that Interaural Cross-Correlation, which is kind of a measure of perceived spatial sound, is affected some by individual HRTF, also positioning, so stereo triangle angle preference could vary some by person and by setup.

Which all reminds that nothing is in isolation, like stereo crosstalk happens with direct sound but also with (some) early reflections, if it matters any.

 

[youngho]

At one time, when Audio Note introduced their speakers it was setup about more than 120 degrees angle by one ‘audiophile expert’. Great reviews about the speakers and the expert’s skills in bring out the best. I must be the only one to say the sound is watered down although the stereo width was at 120 or so degrees but that doesn’t require special skills. Just record one sound in one channel each and you have the width proportional to the speakers placement angles. Some minimalist recordings like the Waterlily acoustics albums sounded nice but Tracy Chapman really sucked. But since the majority praised the system I knew I have serious taste and hearing problem.

I wonder if you could provide a reference for this? I'm curious to read more

The only reliable source why 60 degrees was chosen as the default position was the explanation offered by the eminent recording engineer Robin Miller where he said the mics were placed at 120 degrees and proper level is attained when the speakers angle are half of that.

Do you mean this Robin Miller? http://www.filmaker.com/ I don't understand this explanation as written. Can you please elucidate further? It seems to me that microphone setups and angle can vary, so from what I can tell, ORTF has two microphones angled 110 degrees from each other. Should such recordings be reproduced, then, with speakers subtended at 55 degrees? Blumlein recordings (like some of the Kavi Alexander recordings for Water Lily) and NOS recordings at 45 degrees? Conversely, I mentioned already Robert E Greene discussing 90 degrees subtended angle for Blumlein recordings, echoed here: https://www.psaudio.com/blogs/coppe...j7_mC6iNvhdyUD6_-LqgKFqkuwibWOPnBRTIzKip1TgoU.

On the other hand, Lipshitz mentions in this article previously brought up on ASR (https://www.audiosciencereview.com/...-or-an-objectivist.29115/page-24#post-1024738) that summing error "possibly explains why stereo setups usually a subtended angle of less than 90°--60° is more common, leading to a summing error of just over 1 dB."

 

[Tim Link]

Using the anechoic dummy head data that LION found I did some comparisons using REW. What I did is simply sum the left and right ears for three offset angles; straight ahead to simulate a center speaker, 30 degree offset to simulate the phantom center on an equilateral triangle, and 60 degree offset to simulate the phantom center sound on an extra wide triangle. I'm assuming the time delays between the ears are already included in the impulse files.

And here's what I got:

The normal triangle has it's main dip in response around 2k, while the wider triangle has its main dip down around 700k. Up above 3k till about 10k the standard equilateral triangle is a closer match. Above 10k they both fail but equilateral still does a little better. I can't say from looking at this that my wide setup is a clear winner.

Below is comparisons to center speaker as referenced flat.

Both are altered in the midrange, but differently.

 

[STC]

wonder if you could provide a reference for this? I'm curious to read more

I have tried to justify by giving examples in subsequent posts.

Do you mean this Robin Miller? http://www.filmaker.com/ I don't understand this explanation as written. Can you please elucidate further?

That’s him. I will post here when I find where he said that. It is in one of the Spatial group I think.

What I did is simply sum the left and right ears for three offset angles

I think you are addressing my comment. If so please note I also mentioned why doing summing is not relevant. REW or other measurement can give a lower value due to positive and negative phase summing but at ears it works very differently.

 

[j_j]

I think you are addressing my comment. If so please note I also mentioned why doing summing is not relevant. REW or other measurement can give a lower value due to positive and negative phase summing but at ears it works very differently.

If you want to know the resulting HRTF you do sum them. That cancellation really happens.

 

[LIΟN]

If you want to know the resulting HRTF you do sum them. That cancellation really happens.

Hello j_j. I've been waiting for you since yesterday.
I have done it according to your advice and suggestions. What do you think of this result?

What’s Your Triangle in Stereo Speaker Listening?

The thing to compare is the resulting joint HRTF from both left and right to one ear (either one) to the actual zero degree hrtf. It's an interesting indictment for the phantom center. Thank you for your quick reply. Please understand that I am very overwhelmed to keep up with your knowledge...

audiosciencereview.com

What’s Your Triangle in Stereo Speaker Listening?

The thing to compare is the resulting joint HRTF from both left and right to one ear (either one) to the actual zero degree hrtf. It's an interesting indictment for the phantom center. Thank you for your quick reply. Please understand that I am very overwhelmed to keep up with your knowledge...

audiosciencereview.com

 

[STC]

If you want to know the resulting HRTF you do sum them. That cancellation really happens.

No. If you sum left in and right out of phase you get null. But that doesn’t mean you won’t hear anything. My first response in this thread is only about the XTC part so please read them in that context. If you use Goniometer to check the signal of XTC it would practically show empty display but that doesn’t mean no sound is heard by the listener.

Anyway, I would appreciate the reference or link which you are stated due to HRTF you have difference speed of sound for different frequencies.

 

[j_j]

Anyway, I would appreciate the reference or link which you are stated due to HRTF you have difference speed of sound for different frequencies.

Evaluate the phase of an HRTF. Remove the constant delay part. It's not flat, therefore time of arrival varies with frequency. I doubt anyone's bothered to publish that, it's rather obvious.

 

[STC]

Evaluate the phase of an HRTF. Remove the constant delay part. It's not flat, therefore time of arrival varies with frequency. I doubt anyone's bothered to publish that, it's rather obvious.

I am still searching for the answers which you stated.

While speed of sound varies SLIGHTLY, the irregular shape of the head, etc, causes delay to change a bit at different frequencies. The primary (mean) number is certainly due to the path around the head from the source direction, of course.

But it's not the speed of sound that matters much here, it's the irregular shape of the head and the path from one side to the other.

Your statement is very important to me or anyone interested in crosstalk cancellation because this give a far superior attenuation level.

Now going to the other point which said:-

Evaluate the phase of an HRTF.

This is new again. So what is the phase of HRTF? Or what is HRTF and the purpose?

Remove the constant delay part.

Constant delay of time difference between left and right ear? You mean ITD. But you said they are constant and changes according to frequency. So where does the constant delay is coming from?

It's not flat, therefore time of arrival varies with frequency

What is not flat? Frequencies? I don’t think anyone disputing that. So what else is not flat here?

I doubt anyone's bothered to publish that, it's rather obvious.

I hope at least you could provide a simple
explanation since this is so basic but still beyond my understanding. And since I am the only one asking I am sure others too understood your statement and therefore they could be kind enough share their valuable knowledge.

 

[LIΟN]

What is not flat? Frequencies? I don’t think anyone disputing that. So what else is not flat here?

If what he is talking about varies by frequency, that's true.
i made just dirac pulse. (flat)

Yes. it's flat. and same time.
And i add some And I'll artificially change the frequency as I like. (
PK Fc 4000 Hz Gain 10.00 dB Q 2.000
PK Fc 500.0 Hz Gain -10.00 dB Q 2.000)

Minphase is a change in frequency amplitude, so of course there's a change. I think this is probably what he's talking about. Or maybe I don't know.

can also check it out in the group delay


And to avoid this phase shift, we need to make it linear phase (all delayed to the slowest, symmetrical impulses).
But I'm not sure if that's real and if this is what leads to the comparison of the zero degree and the HRTF sum response at a certain angle we were talking about yesterday.

And I'm a little bit questionable, too.
If you want to observe really pure HRTF, you should use MATLAB to use public sofa and put a pure signal in it.
The public dummy head posted by me (or tim), my thread measurement response, is also not reliable in such a minute delay because HRIR has the influence of the speaker itself.
But the important thing is that I don't know how to handle matlab. lol