• 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!

How do you measure head-located crosstalk?

bachatero

Active Member
Joined
Dec 25, 2023
Messages
253
Likes
194
How do you measure the crosstalk of a speaker system from your ears' perspective? At my desk setup, if I pan music all the way to the left or right and plug my ear that side, I can tell the other side is a lot quieter, but I can't quantify it. Is that possible to do without dropping big bucks on a simulated head and what is a "good" amount? Note: My ideal amount is 0 because I want the "headphones" experience but with speakers, that is perfect soundstage and perfect channel isolation.
 
Last edited:
My ideal amount is 0 because I want the "headphones" experience but with speakers, that is perfect soundstage and perfect channel isolation.
I'm not sure if that's possible and it's not the way we hear real, natural sound, including live music.

...Maybe you should stick to headphones.

I want the "headphones" experience but with speakers, that is perfect soundstage and perfect channel isolation.
Most people don't get a realistic soundstage illusion with headphones. The most common experience is to hear the sound coming from inside the head. (I hear it near my forehead except for hard-panned sounds that I hear coming directly from the headphone transducer into the left or right ear.) But, yes there is total separation and hard-panned sounds will only be heard in one ear (which doesn't happen in "real life").

Tons of sound absorption may help. Floyd Toole mentioned that sometimes speakers in an anechoic chamber give the impression of the sound coming from inside the head, like with headphones.
 
I don't think you understand. Headphones provide you with perfect channel isolation, while speakers provide perfect soundstage due to the sound contacting all your pinna and the low frequencies vibrating your body.
I'm not sure if that's possible and it's not the way we hear real, natural sound, including live music.
Technically, this isn't true because when you listen to recorded music, you aren't listening to real, natural sound, but rather literally a recording. I just want to listen to a recording perfectly and whether that recording represents real, natural sound is up to the recording, and a nonzero crosstalk negatively affects fidelity by definition.
 
perfect soundstage and perfect channel isolation.
These might be contradictory goals as @DVDdoug mentions.

I think if you want very low crosstalk with speakers you need something like BAACH to cancel the crosstalk acoustically. You might be able to devise a DSP setup of your own but I think the head tracking would be a challenge to DIY.

How do you measure the crosstalk of a speaker system from your ears' perspective?
You could place measurement mics at your ears and do it that way?

Or get some in-ear mics?

a nonzero crosstalk negatively affects fidelity by definition.

If this is your position then you should only be using headphones. Zero crosstalk isn't really possible with speakers, even with a lot of effort put into cancelling crosstalk.

However, I would disagree - anything mixed / mastered on speakers was created with a normal amount of crosstalk. So listening to it without crosstalk is effectively changing the format from "the original".
 
You actually don't need total cross talk cancellation. If the cross talk difference is 17 db or great you'll hear it fully to the left or to the right. Many are surprised this is so low. You can prove it to yourself by listening to some filtered noise to find at what level difference you hear it move fully to one side or the other. Since it is possible to hear this my guess is speakers generally have cross talk levels at or a bit below this level depending upon the room. It is mostly set by your head dimensions or HRTF. That is why your hearing works the way it does.
 
However, I would disagree - anything mixed / mastered on speakers was created with a normal amount of crosstalk. So listening to it without crosstalk is effectively changing the format from "the original".
This is tricky because more music is now getting mixed exclusively on headphones because that's where the audience is, and therefore speakers have the same issue where crosstalk changes it from "the original".
 
BACCH uses and sells (not cheap) binaural mics for measuring cross-talk and HRTF for BACCH calibration.

There are much cheaper binaural mics available. See these article by @mitchco who used them for measuring headphones and made binaural recordings of KEF LS50s backed by dual Rythmik L12 subs for his review.
 
This is tricky because more music is now getting mixed exclusively on headphones because that's where the audience is, and therefore speakers have the same issue where crosstalk changes it from "the original".
Sort of goes the other way actually. An old simple mixing technique from the early days of stereo was to put everything equal in both channels, 100% right or 100% left. Old mixing consoles just had a switch for directing tracks left, right or center. Such a recording done now with the high channel separation we have over headphones sounds odd. Over speakers it is fine. The fix for headphones is a little cross feed. Something like that -15 to -17 db of one channel put into the other and it sounds fine over headphones. Headphone mixes that are fine sound fine on speakers too.
 
This is tricky because more music is now getting mixed exclusively on headphones because that's where the audience is, and therefore speakers have the same issue where crosstalk changes it from "the original".
As far as it goes, I don't disagree - but it's not that simple - keep in mind that a lot of headphone-only producers are also using tools like this or this to generate crosstalk (and a simulated room) while they mix. I don't think it's THAT common to produce a true by-and-for headphone mix, but I haven't looked into it very much.
 
Yes you can. If it's for crosstalk, you don't need an expensive microphone. You can buy a cheap 3mm / 6mm binaural microphone.
And you can visualize crosstalk in REW as well, as you can in other programs/tools.

 
I once had the opportunity to measure a popular “image-expander” algorithm. The white paper explained this in terms of crosstalk cancellation. However when I measured this in a real room with in-ear microphones, I found no crosstalk cancellation at all, even though the effect was easily audible. It turned out that the algorithm was increasing the phase difference between left and right channels below 1Khz by forming a filtered L-R signal and adding it to one channel and subtracting it from the other. This only affects signals panned off-center.
I would submit that true crosstalk cancellation in a real room at a distance is almost impossible, and many systems that purport to achieve cancellation are really just playing with low frequency inter-aural phase.
 
I once had the opportunity to measure a popular “image-expander” algorithm. The white paper explained this in terms of crosstalk cancellation. However when I measured this in a real room with in-ear microphones, I found no crosstalk cancellation at all, even though the effect was easily audible. It turned out that the algorithm was increasing the phase difference between left and right channels below 1Khz by forming a filtered L-R signal and adding it to one channel and subtracting it from the other. This only affects signals panned off-center.
I would submit that true crosstalk cancellation in a real room at a distance is almost impossible, and many systems that purport to achieve cancellation are really just playing with low frequency inter-aural phase.
you mean mid side things-image or stereo expander?
 
I once had the opportunity to measure a popular “image-expander” algorithm. The white paper explained this in terms of crosstalk cancellation. However when I measured this in a real room with in-ear microphones, I found no crosstalk cancellation at all, even though the effect was easily audible. It turned out that the algorithm was increasing the phase difference between left and right channels below 1Khz by forming a filtered L-R signal and adding it to one channel and subtracting it from the other. This only affects signals panned off-center.
I would submit that true crosstalk cancellation in a real room at a distance is almost impossible, and many systems that purport to achieve cancellation are really just playing with low frequency inter-aural phase.
This was actually something from Alan Blumlein in the 1930s when he developed his idea for stereo. He called it a shuffler circuit. It used filtered mid-side processing to do this.

Here is a Michael Gerzon paper about it from 1986.

There are some systems that do some crosstalk cancellation. Polk used to make speakers with extra drivers in each channel just to cancel some of the other channel. Carver had it in some of his preamps. There are software implementations around of course.
 
you mean mid side things-image or stereo expander?

Probably SRS. They did use Shuffler effect and so too some XTC use shuffler to adjust the centre and width by increasing the ILD. I never understood how the changes using Shuffler effect helps the colouration unless it is more than simple M/S processing.
 
Last edited:
Probably SRS. They did use Shuffler effect and so too some XTC use shuffler to adjust the centre and width by increasing the ILD. I never understood how the changes using Shuffler effect helps the colouration unless it is more than simple M/S processing.

I don't know if he's talking about anything other than what you said.
But if he's just talking about XTC as a regular mid/side (through the channel mixing concept), it's kind of a puzzle



1725361705517.png


1725361982074.png


When intentionally combined like this, the intersection where the information of the two channels overlaps is the mid, and the rest of the information except for the mid is the side.
In the end, L+R or R+(-L_)
Mid + Side
Mid + (-Side)
It's just this channel mixing.


But no matter how much you adjust the mid and side of the music stage with channel mixing, it ends up in our two ears.

1725362107010.png

1725362130009.png



So the perceived mid/side feel will of course change, but I think it's just another music playback with channel mixing because the underlying face and opposite ear effects still remain.
That's why I think the author who wrote the mid/side article confuses the part related to the sound source and the part related to the intermediate DSP processing.
 
  • Like
Reactions: STC
I don't know if he's talking about anything other than what you said.
But if he's just talking about XTC as a regular mid/side (through the channel mixing concept), it's kind of a puzzle



View attachment 390044

View attachment 390045

When intentionally combined like this, the intersection where the information of the two channels overlaps is the mid, and the rest of the information except for the mid is the side.
In the end, L+R or R+(-L_)
Mid + Side
Mid + (-Side)
It's just this channel mixing.


But no matter how much you adjust the mid and side of the music stage with channel mixing, it ends up in our two ears.

View attachment 390046
View attachment 390047


So the perceived mid/side feel will of course change, but I think it's just another music playback with channel mixing because the underlying face and opposite ear effects still remain.
That's why I think the author who wrote the mid/side article confuses the part related to the sound source and the part related to the intermediate DSP processing.
XTC is mid side processing. ✍️ meant to say not mid side processing. Spatial effect and XTC are not the same.
 
Last edited:
XTC is mid side processing. Spatial effect and XTC are not the same.
You're probably talking about the difference between the channel mixxing effect and XTC, right?
Mid/Side allows you to highlight or reduce certain information by summing and subtracting between both channels, but that doesn't lead to cancellation (because listening with both ears is the same). The process of mixing both channels looks similar, but the timing and implementation process are very different, and the two purposes are different in the first place. I think you probably agree with me, but I think it's an error in translation
 
Yes I was talking about SRS (and the many variations people used to get around the original patent). But I would be interested to see if any of the crosstalk cancellers really work (as verified by in-ear mics) in an actual room without the listener’s head being bolted to a stationary object. If the impulse response at each ear shows a clear dominant arrival, with the clutter of reflections arriving later (10-20 ms) then XTC might give you want you want even though the RMS value of the response at the “cancelled” ear is not zero, because perceived directionality relies on that first arrival.
 
Last edited:
Back
Top Bottom