Stereo Crosstalk Elimination (reduction) Par Excellence!

[Tim Link]

I feel like I've come across something really fantastic! It started out as an attempt to simply derive a center channel from 2 channel stereo. By accident it turned out to be an outstanding crosstalk elimination method - much better than I thought at first because I hadn't dialed it in all the way. Turns out this sounds freakin' beautiful when listening on-axis, and still retains good tonal quality when off axis or even listening from another room. A slight adjustment transformed it from merely interesting to stunning.

For a long time now I've been interested in reducing stereo crosstalk that occurs from the sound of the speakers reaching both ears instead of just the intended ear. The gold standard for me was the result from using a physical divider that went right up to my face. Sounded amazing but a really uncomfortable way to listen. Now I've managed to get stellar results with no divider wall required, and it can be done with 3 small speakers and a two channel amplifier, with some very simple conversion of the left and right channels by converting one channel to sum and the other to difference. I did this in the digital domain on my Mac mini using Audio Hijack, as shown below.

The really important tweak to this Audio Hijack arrangement was adding the balance box at the end, which now instead of making the sound shift left or right, makes it wider or narrower by changing the ratio of the side channels to the center channel. You could also do this with the balance control on your pre-amp if it has one. The -12 percent off the center channel did the magic, fixing the somewhat hollow tone that was occurring and just bringing everything to life.

You do need three identical small speakers. They need to be arranged all in a row close enough together so the tweeters are ear distance from one another to get optimal on-axis response. The speakers must be narrow enough to accommodate that.

The speakers are wired with the center speaker getting the summed mono signal - which is the amp's right channel in my arrangement, and then the two outside speakers get the amp's left channel. I have them wired in parallel, with the right side speaker wired up backwards. This gives a L+R signal to the center speaker, a L-R signal to the left speaker, and a R-L signal to the right speaker.
Below you can see my humble array of three Sony SS-CS5 speakers which are currently blowing my mind. This is a different take on the Polk SDA concept and I think it works better.
If you're at all interested in crosstalk elimination I highly recommend you give this arrangement a try. This produces a very pure, clean, detailed, non-gimmicky stereo sound with excellent tone and musicality. No heavy handed processing is involved. Just simple summing and difference signals.

 

[tmuikku]

Hi, here is similar arrangement: http://elias.altervista.org/html/SingleSpeakerStereo.html
Although your method works with classD amps as well as grounds are not tied together like Elias'

 

[Tim Link]

Hi, here is similar arrangement: http://elias.altervista.org/html/SingleSpeakerStereo.html
Although your method works with classD amps as well as grounds are not tied together like Elias'

Thanks! Very interesting link. He obviously was thinking along the same lines, but as I initially did, he missed the opportunity for crosstalk cancellation by time aligning the drivers for each ear. I know not everybody likes crosstalk cancellation, and not everybody likes a solid center image, preferring the ghost or phantom image because they feel it provides a better sense of depth. I think it does sometimes, depending on the recording.

 

[sarumbear]

Interesting to read stereo to be defined as crosstalk

 

[Tim Link]

Interesting to read stereo to be defined as crosstalk

Stereo can mean any number of speakers beyond mono. 2 channel stereo is the simplest stereo and it works really well, as we all know. But, a lot of inter-aural crosstalk occurs with just two speakers. The crosstalk is strongest right in the middle of the sound stage and goes to zero if any sound is hard panned to one speaker or the other. As Amir has pointed out, we can hear past the crosstalk interference and other comb filtering caused by wall and ceiling reflections so it still sounds really good to us. But you can hear a difference if the crosstalk is reduced. The stereo effect widens and clarifies and the tonality changes as a result of the reduced comb filtering. I've not tried it by I've read that listening to 2 channel stereo in an anechoic space will reveal a marked tonal shift as a vocal image is panned from left or right to center. In a reflective room this is largely obscured, but not totally.

 

[sarumbear]

Stereo can mean any number of speakers beyond mono.

I can only repeat what I said above. You have an interesting definition. The term stereo was accepted as a two channel format throughout all my adult life and audio education. That is why there exists the word “stereo symmetry” that defines, for instance, the human hands

 

[Tim Link]

I can only repeat what I said above. You have an interesting definition. The term stereo was accepted as a two channel format throughout all my adult life and audio education. That is why there exists the word “stereo symmetry” that defines, for instance, the human hands

Agreed, since 2 channel has dominated for decades, the word stereo and 2 channel have become virtually synonymous. To differentiate higher numbers of channels for marketing purposes new terms have been required. We have stereo hearing involving two ears oriented in a horizontal way. Owls have some vertical orientation as well by aiming each ear in a different direction. The means of reproducing a sound field for stereo hearing can involve any number of speakers. https://en.wikipedia.org/wiki/Stereophonic_sound

 

[Tim Link]

My growing family of cheap Sony speakers! More speakers in the array adds recursive crosstalk reduction, resulting in a more pure and detailed sound stage with more solid and precise imaging in the areas far to the left and right of the speakers themselves. The sound field can be ridiculously widened by turning down the center speaker, causing center panned sounds to seem more distant and sometimes reverberant if the reverb was not panned center. I'm still not sure where the center level should be set, and I think some EQ or bass management is called for. This definitely improved the stereo imaging but the tonality needs to be worked out again.

 

[MCH]

I can only repeat what I said above. You have an interesting definition. The term stereo was accepted as a two channel format throughout all my adult life and audio education. That is why there exists the word “stereo symmetry” that defines, for instance, the human hands

While I agree with what you are saying, stereo in symmetry or in chemistry, where it is widely used, refers more to "space" or "3d" if you want than to pairs of two. You can have more than two structures that are stereoisomers of each other. A particular case of stereoisomers are enantiomers (specular images, like your hands in your example, that are logically only two possible).
I don't know how the word stereo made it to audio, but could well be that it was referring to something like "3D sound" rather than 2 sources.

 

[tmuikku]

Thanks! Very interesting link. He obviously was thinking along the same lines, but as I initially did, he missed the opportunity for crosstalk cancellation by time aligning the drivers for each ear. I know not everybody likes crosstalk cancellation, and not everybody likes a solid center image, preferring the ghost or phantom image because they feel it provides a better sense of depth. I think it does sometimes, depending on the recording.

Hi, do you mean that he doesn't mention about it and doesn't utilize it? (the crosstalk cancellation, combfilter)

When the source is in front, there is no, or very short, path length difference to each ear and so not much combfilter from the direct sound. Path length difference to each ear gets greater as the source is off center and is greatest when source is 90deg to side of head. In this regard I think your and Elias' solution is similar (enough), sources are practically front. Or did I miss something?

For stereophonic sound we need different sound to each ear, and in your and elias' case its reflections from each side wall that do it? Is the difference here, instead of pointing speakers to walls point all speakers toward listener along with the matrix so that sound through each boundary is similar for each speaker and now contains some of the crosstalk cancellation features?

Could you elaborate more on what you think is happening, thanks!

ps. tip for the mid/side balance I believe there is no right or wrong, as listener you are supposed to enjoy it so adjust each song to taste If in doubt have it more mono, I had single speaker mono setup for few years and it was mighty fine.

 

[tmuikku]

Thinking of your system, so it works the side speakers are physically ear distance apart to cancel each side out to the other ear, right? In which case the scheme should work best when L and R are next to each other, to get their distance the ear distance, ~17cm, 6-7"?

I suspect you have arranged the speakers physically so that center sound is the center one (L-C-R) as its logical. Have you tried the L and R speakers next to each other (C-L-R)? You could have the center channel speaker sit above the two sides, upside down to get all tweeters close to ear height.
C
L-R

 

[charleski]

Hi, here is similar arrangement: http://elias.altervista.org/html/SingleSpeakerStereo.html
Although your method works with classD amps as well as grounds are not tied together like Elias'

elias did a 3-speaker config as well:

http://elias.altervista.org/html/3_speaker_matrix.html

There have been various DIY implementations of this that you can find, such as this one, there's an interesting description of the subjective effects further down the thread. Unfortunately I've never heard a setup like this.

Commercial efforts seem mostly aimed at 'one-speaker stereo' offerings such as:

SPACESTATION V.3 – Aspen

aspenpittmandesigns.com

I suppose I have to speculate that the cheap Sony speakers you're using are probably not the greatest performers in a classical stereo config. But it would be interesting if this were a way to 'rescue' a decent soundfield from a bunch of cheap 'n' nasty HT speakers.

 

[gitaarwerk]

Out of curiosity, how is this different than Mid/Side processing? I used this with mixing and playing with modular synthesizer, process the sides or the mids only for some weird effects; Not put the separate channels as different speakers, but could do that as well. I'm not per se sure if this makes it easier. Probably easier to get results; but not sure if they are better or worse. Every time you add something, it gains complexity in some domain. Just thinking aloud. Nevertheless, it's an interesting way to do music. Rather than go for quadraphonic and up, just go sides + center.

 

[tmuikku]

^^ Yeah these single structure stereo systems are all more or less the same concept. Three speaker configuration like normal stereo + center is bit different, but attacks the same issue, combating comb filter of the center sound, replacing phantom center with real center. I've build (and since scrapped) the elias single speaker stereo and it worked nice, except there was difference between the direct and side sound in a way that although there was nice mono and nice "canvas" like sound field on the wall I had the speaker against, it was somehow disconnected. Perhaps I had some issues with the build that are not directly related to the configuration or something.

Perhaps, if the tweeters were all pointed towards the listener, like here in this thread, the direct sound would have least amount of interference to it, cancelled by hearing system even if it works as advertized. I would imagine this works best with multiway speakers and not as good with fullrange drivers (like elias prototype) that beam the highs. I think there is need to have the sound splashed to side walls while having coherent direct sound which should both happen in this thread concept, if I'm imagining it right. Anyway, this could work bit better than the other schemes that splash the sidewalls and try avoid corrupting the direct sound failing at it

 

[tmuikku]

... Probably easier to get results; but not sure if they are better or worse. Every time you add something, it gains complexity in some domain. Just thinking aloud. ...

Yeah even in single domain, acoustics. sound propagates spherically, to all directions. If you think about it there is no way two sound sources that are not physically at same location would perfectly cancel out to every direction so there is always some "residual", worse interference to some directions and varying with wavelength. And there is no need to because you just use one source and then manipulate it electronically For example the Polk SDA system mentioned uses similar crosstalk cancellation as here, two sources spaced ears apart but what I see it fails, like any other schemes, is that also the "cancelling" sound arrives both ears so its just replacing crosstalk with another interference, perhaps different but the problem of having ears each side of head still persists. As I understand it the underlying issue is that sound panned to center comes from multiple different physical locations, which both arrive to both ears at slightly different time causing comb filter. Three speaker stereo setup has single source for center sound which ought to fix most of the issue leaving slightly panned sound still somewhat affected by the issue.

We have two ears and hearing system and we can exploit it as well as possible for enjoyable sound of course, but there is no escape the fact that sound goes around head and reaches both ears, after all this is what allows localization in the first place Except with headphones sound goes from one source to one ear, similarly with the separator wall also mentioned in the first post. One could use any of these methods, what ever they fancy and feels the best. If it has to be speakers then best would be to have multiple sound sources and multichannel mixes so that each instrument has their own speaker, like in reality each performer and sound source are located at different, real, physical locations. This would be cumbersome and expensive niche. Sometimes mono / single package is the ticket, and having good sound from it, with stereo effect even, would be nice.

 

[tmuikku]

Free thinking about it: why not have open back headphones playing the sides signal, no crosstalk at all, and then single speaker mono somewhere in the room to get the sound out of head? Augment with face tracking to adjust headphone delay with distance to the mono speaker. Or just add some crossfeed or HRTF stuff to headphones to have the perfect situation and get the sound out of head? What are we chasing at here, our own tail? Different systems for different applications, anyone can choose that fits for them and the other perspective is being a business trying to sell stuff to people. How about horizontal speaker array, signal somehow spread on it so that at any given time only one or few adjacent transducer output the same sound, 30 speaker stereo instead of 3. All right, getting out of topic, but thinking stuff is fun

 

[tmuikku]

My growing family of cheap Sony speakers! More speakers in the array adds recursive crosstalk reduction, resulting in a more pure and detailed sound stage with more solid and precise imaging in the areas far to the left and right of the speakers themselves. The sound field can be ridiculously widened by turning down the center speaker, causing center panned sounds to seem more distant and sometimes reverberant if the reverb was not panned center. I'm still not sure where the center level should be set, and I think some EQ or bass management is called for. This definitely improved the stereo imaging but the tonality needs to be worked out again.View attachment 243666

So the crosstalk cancellation system works as long as the setup is physically symmetric sides have signals L-R and R-L? spacing doesn't have to be distance between ears?
What if you spread your 5 speaker array wider as wide as you like? What if you gave 0.5*L-0.5*R and 0.5*R-0.5*L to the speaker between extremes and center, or how to further split the width into individual sets of speakers or is there any point doing so? Can you just leave out the in-between speakers and have three spaced out? What about distance, perhaps speakers should arc to maintain path length and clarity? Just thinking out loud

edit. well, perhaps I should get reading papers before posting anymore http://decoy.iki.fi/dsound/ambisonic/motherlode/source/7036.pdf
Carry on

 

[Tim Link]

Hi, do you mean that he doesn't mention about it and doesn't utilize it? (the crosstalk cancellation, combfilter)

When the source is in front, there is no, or very short, path length difference to each ear and so not much combfilter from the direct sound. Path length difference to each ear gets greater as the source is off center and is greatest when source is 90deg to side of head. In this regard I think your and Elias' solution is similar (enough), sources are practically front. Or did I miss something?

For stereophonic sound we need different sound to each ear, and in your and elias' case its reflections from each side wall that do it? Is the difference here, instead of pointing speakers to walls point all speakers toward listener along with the matrix so that sound through each boundary is similar for each speaker and now contains some of the crosstalk cancellation features?

Could you elaborate more on what you think is happening, thanks!

ps. tip for the mid/side balance I believe there is no right or wrong, as listener you are supposed to enjoy it so adjust each song to taste If in doubt have it more mono, I had single speaker mono setup for few years and it was mighty fine.

He's not utilizing stereo crosstalk cancelation because he doesn't have the speaker drivers properly aligned for that.

It is true that there is no comb filtering from the center channel by itself. There is, however from the side channels, both with each other and with the center channel as they are all reaching each ear at a slightly different time in Elias' arrangement. With my three speaker setup the center channel reaches each ear at the same time as the corresponding side channel reaches that ear, so the left side channel reinforces the left signal coming from the center channel at the left ear and cancels the right signal at the left ear, so ideally the left ear hears only the left channel. The same thing happens on the other ear so both ears hear mostly just the channel they're supposed to hear. But then some of the right side channel leaks in to the left ear so it's not perfect. For the channel reinforcement and cancellation to work, the drivers need to be ear distance apart and all in a row in front of or behind the listener. With that arrangement the crosstalk cancellation and side channel reinforcement will work correctly over a range of listening distances from the speakers. To cancel the remaining side channel crosstalk you can just add another set of side speakers, again placed ear distance apart further out, playing the same side channel information. That cancels the first set of crosstalk signals from the opposite side speaker but creates a new crosstalk signal that is a bit weaker and more delayed. You can add side speakers indefinitely to weaken and further delay the crosstalk but I think 2 on each side of the center for a total 5 total speakers is enough to produce an amazingly clear and wide stereo sound field. My speakers are actually a bit too wide. The drivers should be slightly closers together for my ear spacing but it works nonetheless.

To elaborate more - my system, unlike Elias', has nothing to do with sidewall reflections. It's all about the direct sound reaching your ears, reinforcing the correct side channel information at the correct ear and canceling the wrong side channel information. If perfected, each ear hears only what it is supposed to with no leakage across the head. Fortunately it doesn't have to work perfectly to produce excellent imaging. It's kind of like headphones except the sound is traveling from front to back across your head and pinnae like it's supposed to, so it's highly externalized - out in front of you. This will work perfectly well in an anechoic chamber in terms of producing a very wide, clear, and well fleshed out sound field that is externalized.

I think you are correct about the mid / side balance being somewhat of a preference thing. I've found a setting that I like well enough for all kinds of recordings now, but I've noticed that I can get away with a little less mid channel on big orchestral works. It pushes the sound stage back and makes for a grand, lush presentation. If you try that on some vocal mixes the lead singer can get sort of lost sounding, like they're singing from down the hallway while the instrumental backup is in the room with you.

 

[Tim Link]

Thinking of your system, so it works the side speakers are physically ear distance apart to cancel each side out to the other ear, right? In which case the scheme should work best when L and R are next to each other, to get their distance the ear distance, ~17cm, 6-7"?

I suspect you have arranged the speakers physically so that center sound is the center one (L-C-R) as its logical. Have you tried the L and R speakers next to each other (C-L-R)? You could have the center channel speaker sit above the two sides, upside down to get all tweeters close to ear height.
C
L-R

The speakers need to be ear distanced apart with the center speaker directly ahead of you. That puts your ear at the half way point between the center driver and the side driver, so the sound from the center and the side reach the corresponding ear at exactly the same time. Since the center speaker is playing L+R and the side (let's suppose left side) is playing L-R, you end up with 2L and 0R signal strength at the left ear, which is just two left signals, or the left signal elevated by 3dB and the right signal obliterated. It's all perfect but then the right side channel leaks in to your left ear with a R-L, so now we'd be back to L+R in each ear again, which is mono sound. But fortunately the leaking right channel signal is not time aligned with the original center channel signal at the left ear so it doesn't reinforce or cancel so neatly, so it's ability to erode the stereo sound field is greatly diminished.

With a regular 2 speaker stereo arrangement you also get L and R signal at both ears, with the opposite signal slightly delayed and attenuated by traveling across the head at an angle. There's no beneficial canceling or reinforcing so the stereo effect is not as clear but still it's enough for us to hear a decent stereo image. What I think is troublesome about it is that the best coherent sound from a standard setup comes from hard panned left and right signals. The phantom center image has the most crosstalk and is the most compromised sound, but this is usually where the focal point singer or instrumentalist is located. Also, the sound that is supposed to be coming from straight ahead is actually traveling sideways across our head from two different directions. It's a bunch of little issues that we hear through but it's not ideal even with the best 2 channel arrangements. With my system the straight ahead sounds cross the head correctly but sounds panned far to the right or left are now still actually coming from more or less straight ahead. So we've traded one problem for another but since I consider the center of the sound field to be more important I think this is a good trade.

 

[Tim Link]

elias did a 3-speaker config as well:

http://elias.altervista.org/html/3_speaker_matrix.html

There have been various DIY implementations of this that you can find, such as this one, there's an interesting description of the subjective effects further down the thread. Unfortunately I've never heard a setup like this.

Commercial efforts seem mostly aimed at 'one-speaker stereo' offerings such as:

SPACESTATION V.3 – Aspen

aspenpittmandesigns.com

I suppose I have to speculate that the cheap Sony speakers you're using are probably not the greatest performers in a classical stereo config. But it would be interesting if this were a way to 'rescue' a decent soundfield from a bunch of cheap 'n' nasty HT speakers.

I think this really is a way to rescue a descent sound field from a bunch of cheap 'n' nasty HT speakers. As long as they're small enough to get close enough together they should produce an excellent sound field. Tonality and distortion issues remain. The little Sony speakers are making an amazingly big sound now that I've got the setup balanced out better. I took it to work today and set it up in our listening room. It stomps the crap out of the much larger KRK monitors that are in there in terms of producing a big, deep sound field, and in terms of overall fun factor. I finally got one of my co-workers to sit down and have a listen. On first entering the room he thought it sounded a bit strange, but once he got seated between the speakers he heard it open up and noted that it produced a big sound field and some "serious separation" between the singers. We were listening to Weaver's Reunion at Carnegie Hall. He pointed right to each singer and each instrument in the back ground. I was standing right behind him and from where he was pointing he was hearing it just like me. It's freakin' impressive, and the images are very solid and rich and clearly located even out to the corners of the room. Nothing ghostly or phantom like about them.

Elias' posts are very interesting and I may have to try those after reading more thoroughly to understand how to implement those concepts. Maintaining proper energy at each frequency is a complex problem and I haven't tried to mathematically understand it yet but I know it's important. All the imaging in the world is for naught if the tone is off. Getting the tone reasonably balanced brings it to life.

The subjective report of Elias' arrangement is interesting. It reminds me of what I hear if I take my three speakers and spread them apart and walk around the room. It sounds really good in some ways, definitely listenable. And indeed, reverb in the recording seems enhanced and you hear things pop out differently than in a normal configuration.

That space station thing is a trip! I wonder what they've got going on inside there. It certainly doesn't look like the kind of thing that would produce amazing stereo.