Now That Atmos Is Everywhere… Real vs. Phantom Center in a 5.1 Music-Focused Setup

[moonlight rainbow dream]

Most of the Atmos mixes I’ve listened to don’t use the center channel for much more than a few elements like the kick and the snare drum, which in most cases are also panned to the front left and right channels as well.

The vocal track is usually panned as a phantom sound, just like it's done for ordinary stereo mixes. I would say that is true for about 90% of all Atmos mixes I have heard.

This is my experience as well dabbling a bit into Atmos music with Amazon. It's sad, because I think the center channel is one of the best parts of multichannel audio. I may just resign myself to upmixing stereo using DSU.

 

[sfogg]

Good processing can work very well for 2 channel stereo music to surround. And some give a lot of controls so you can set the perspective the way you want it.

If you stick to music from 5.1 sources at least a few processors can also perform steering on L/R to use a center as well as bring out more surround info from L/R to expand 5.1 to 7.1. My Lexicon MC-12HD can do that (5.1 Logic 7) and I did it with 5.1 surround music that didn't really use the center. It could accept a 7.1 source too but wouldn't perform surround processing on it.

My QLI-32 can take 2,5.1 or 7.1 and it can expand that up to 32 channels. It works very differently though as it basically breaks the source down into streams and then isolates the direct sound, first reflections and reverb from each stream.

 

[tengiz]

The whole "center channel fixes stereo" line is mostly a pop/film bias. If you’re mixing Top-40 or dialogue, sure - fixing the lead in a real speaker makes life easier. But in opera, symphony, chamber music? There’s rarely a dead center source to begin with. Soloists are off to the side of the conductor, concertos are offset, and the stage itself is designed around asymmetry.

So in those genres the phantom center isn’t the "big problem" it’s made out to be - it’s often irrelevant. Even with the real center, the EQ quirks and comb filtering are still there, just attached to different pan positions depending on where the performers actually sit. Which is why a discrete center may be a godsend for movies and pop (until it isn't when the next best thing comes along), but in classical it doesn’t really fix anything. The live reference itself is wide and lopsided.

I get that the dollar signs shine the brightest in pop/films, but excuse me if I don’t swallow the marketing pitch whole.

 

[sfogg]

The whole "center channel fixes stereo" line is mostly a pop/film bias. If you’re mixing Top-40 or dialogue, sure - fixing the lead in a real speaker makes life easier. But in opera, symphony, chamber music? There’s rarely a dead center source to begin with. Soloists are off to the side of the conductor, concertos are offset, and the stage itself is designed around asymmetry.

The EQ quirks and comb filters are not still there to anywhere near the same amount. In the real world, we hear vocals originating from one point in space (heard from two points) literally all the time. A vocal coming out of a center channel is dramatically closer to that than a vocal originating from two widely spaced points in space and being received by two points in space. The geometry is different even if perfectly centered between L/R and it only gets worse if you aren't centered or not facing dead ahead.

So in those genres the phantom center isn’t the "big problem" it’s made out to be - it’s often irrelevant. Even with the real center, the EQ quirks and comb filtering are still there, just attached to different pan positions depending on where the performers actually sit. Which is why a discrete center may be a godsend for movies and pop (until it isn't when the next best thing comes along), but in classical it doesn’t really fix anything. The live reference itself is wide and lopsided.

The live reference has a vocal originating from one point in space. Doesn't matter where they are, it is still one point. Just lots of different single points in space for multiple people.

For a vocalist panned somewhere off center you will still get comb filtering (even with a center speaker) but you have now cut the angle between the two sources in half. That makes it more stable to listener position and still reduces the comb filtering compared to just L/R.

I get that the dollar signs shine the brightest in pop/films, but excuse me if I don’t swallow the marketing pitch whole.

But you have... for stereo.

 

[tengiz]

The EQ quirks and comb filters are not still there to anywhere near the same amount. In the real world, we hear vocals originating from one point in space (heard from two points) literally all the time. A vocal coming out of a center channel is dramatically closer to that than a vocal originating from two widely spaced points in space and being received by two points in space. The geometry is different even if perfectly centered between L/R and it only gets worse if you aren't centered or not facing dead ahead.

Worse for what? For comb filtering? Didn’t you basically say the opposite in your first sentence? And why does it matter if I’m perfectly fine with the ultimate egoist setup, always sitting in the best possible position for stereo?

The live reference has a vocal originating from one point in space. Doesn't matter where they are, it is still one point. Just lots of different single points in space for multiple people.

How is this relevant to reproduction? A single point source in the original may well be best reproduced from a single center speaker, but that only applies to sources that actually sit in the dead center. Every other source in reproduction comes from two speakers - there’s no way around it.

And how does the fact that the original was a single point eliminate comb filtering? A phantom panned off-center with L/C/R may reduce it somewhat compared to a dead-center phantom for L/R, but it doesn’t suddenly - poof! - disappear. I get that adding a C makes things somewhat better in some cases, but it does not solve the problem.

For a vocalist panned somewhere off center you will still get comb filtering (even with a center speaker) but you have now cut the angle between the two sources in half. That makes it more stable to listener position and still reduces the comb filtering compared to just L/R.

Depending on the geometry, cutting the angle in half by itself may not make much difference for comb filtering. What really matters are the differences in propagation delays between speakers and ears. The result of the angular symmetry - in terms of propagation - is that you can end up with the same delays even at very different angles. I’m sure I’m not saying anything you don’t already know.

In any case, in my experience I hardly even notice it.

But you have... for stereo.

I have, and I am happy with it because it works for me just fine. And I am also happy for those who find physical center better for what they consider higher priority scenarios for themselves.

 

[sfogg]

Worse for what? For comb filtering? Didn’t you basically say the opposite in your first sentence? And why does it matter if I’m perfectly fine with the ultimate egoist setup, always sitting in the best possible position for stereo?

Because even in the best possible position you have 2 sources to two different ears. The timing is different than what occurs in the real world. If a vocal is front and center in a recording (often) then a center gets you much closer to the real world. If they are off center it still helps and is an improvement compared to phantom.

How is this relevant to reproduction? A single point source in the original may well be best reproduced from a single center speaker, but that only applies to sources that actually sit in the dead center. Every other source in reproduction comes from two speakers - there’s no way around it.

Not every source, but most do. But now you have three points anchoring the imaging instead of two. More up front would make it better still.

And really if we are talking surround you have more than 3 points. The sides/rears/heights can add cues that reinforce the imaging up front too. Front heights seem to as well, which I was not expecting. The difference in layering of the front stage was suprising.

And how does the fact that the original was a single point eliminate comb filtering?

What is a single point source going to comb filter with? For comb filtering (wave interactions) you need multiple waves from different locations.

A phantom panned off-center with L/C/R may reduce it somewhat compared to a dead-center phantom for L/R, but it doesn’t suddenly - poof! - disappear. I get that adding a C makes things somewhat better in some cases, but it does not solve the problem.

Better in all cases. Never said it solves the problem completely, said it is better than 2 channel.

Depending on the geometry, cutting the angle in half by itself may not make much difference for comb filtering. What really matters are the differences in propagation delays between speakers and ears. The result of the angular symmetry - in terms of propagation - is that you can end up with the same delays even at very different angles. I’m sure I’m not saying anything you don’t already know.

You aren't considering the two ear part of this problem. If you cut the angles in half you have reduced the interaural time difference (ITD) between each source and EACH of your ears. It isn't propagation delay that we are talking about, it is the ITD.

 

[tengiz]

Because even in the best possible position you have 2 sources to two different ears. The timing is different than what occurs in the real world. If a vocal is front and center in a recording (often) then a center gets you much closer to the real world. If they are off center it still helps and is an improvement compared to phantom.

How does it help with symphonic recordings? Or with opera, where multiple performers are moving across the stage? A center speaker can lock one source, sure - but in those cases the whole point is a wide and/or shifting soundstage. What exactly is the center "fixing" there?

Not every source, but most do. But now you have three points anchoring the imaging instead of two. More up front would make it better still.

And really if we are talking surround you have more than 3 points. The sides/rears/heights can add cues that reinforce the imaging up front too. Front heights seem to as well, which I was not expecting. The difference in layering of the front stage was suprising.

If you’re building a case for going from 2 to 20 speakers across the front hemisphere, then I don’t disagree in principle

But I’m doubtful that going from 2 to 3 makes a huge difference for classical, choral, or anything with multiple distinct, equally important sources spread left to right - especially when there’s nothing of critical importance parked in the dead center.

What is a single point source going to comb filter with? For comb filtering (wave interactions) you need multiple waves from different locations.

Yes, and I've never said otherwise.

You aren't considering the two ear part of this problem. If you cut the angles in half you have reduced the interaural time difference (ITD) between each source and EACH of your ears. It isn't propagation delay that we are talking about, it is the ITD.

Of course I am - the propagation delay I mentioned already accounts for both the spacing between the speakers and the spacing between the ears.

 

[sfogg]

How does it help with symphonic recordings? Or with opera, where multiple performers are moving across the stage? A center speaker can lock one source, sure - but in those cases the whole point is a wide and/or shifting soundstage. What exactly is the center "fixing" there?

If is solidifying the aspects that are the same between L/R. You can't image partway between L/R and not have some overlap between channels. Instead of imaging strictly from L/R it can make it so that the imaging is more between L and C or C and R. It depends upon how it is mixed (if multichannel source) or how the center is derived if a stereo source.

Again, it is essentially cutting your phantom imaging angle in half. Good use of a center channel isn't an all or nothing affair.

I would also think it helps with other more esoteric things such as IMD, dynamics, (potentially dynamic compression) and maybe THD since you now have three speakers working to produce the same acoustic output as two. For some systems that might be a benefit, for others it might not be. And if the center can't keep up than that could be a problem too.

If you’re building a case for going from 2 to 20 speakers across the front hemisphere, then I don’t disagree in principle

But I’m doubtful that going from 2 to 3 makes a huge difference for classical, choral, or anything with multiple distinct, equally important sources spread left to right - especially when there’s nothing of critical importance parked in the dead center.

Clrearly you are doubtful. I'm disagring with you. In my experience, it helps with all of that. *IF* done well.

But that caveat is no different then pointing out a Klipschorn for the left speaker and a Minimus 7 for the Right speaker isn't a good demonstration of what stereo can do either.

The problem, for some, is the elimination (or at least reduction for off center) of some of these comb filtering problems is a change from 'how it sounds' on stereo so it gets rejected out of hand. After listening for awhile one might realize it sounds more real. Literally, you can do something better without it also being different. Listen long enough and going back to phantom imaging sounds kind of hollow.

Of course I am - the propagation delay I mentioned already accounts for both the spacing between the speakers and the spacing between the ears.

Add a center equidistant (or time aligned) with your L/R and the propagation delay hasn't changed between a phantom center image and the actual center playing. The ITD has though.

Unless you are talking about propagation delay from Left speaker to Left ear, left speaker to right ear, right speaker to left ear and right speaker to right ear. Your brain uses ITD to help localize sounds along with HRTFs. Phantom imaging screws with those mechanisms because we are getting conflicting info.

 

[tengiz]

If is solidifying the aspects that are the same between L/R. You can't image partway between L/R and not have some overlap between channels. Instead of imaging strictly from L/R it can make it so that the imaging is more between L and C or C and R. It depends upon how it is mixed (if multichannel source) or how the center is derived if a stereo source.

Thus making the situation of the sweet spot even more fragile - now I need to worry about sitting in the double sweet spot and if I move left or right I ruin both.

Unless you are talking about propagation delay from Left speaker to Left ear, left speaker to right ear, right speaker to left ear and right speaker to right ear. Your brain uses ITD to help localize sounds along with HRTFs.

I am.

Clrearly you are doubtful. I'm disagring with you. In my experience, it helps with all of that. *IF* done well.

As I said, I don't disagree with adding more physical speakers in principle, I just don't see much payoff - for myself - in going from two in the front to five or so. But going from five to hundreds of individually controlled vertical line sources is a whole different ball game as long as they are spaced tightly enough to avoid diffraction and lobing at the wavelengths that matter. It is an old idea, but still mostly in the research domain.

Your entire front wall covered with narrow vertical speaker strips, maybe a hundred per meter. Each strip is individually driven, so the wall becomes a phased array. The wall can launch flat sound waves from any horizontal angle, combine multiple ones at once, and essentially program the sound field. No more fragile sweet spots, no more comb filtering, no more relying on phantom images holding together - just stable, controllable acoustics. Finally, the actual, real object oriented solution that is what it says it is.

That is why, when this trickles down to consumer scale (even in a cut-down form), it will not be just another incremental upgrade. It will be a qualitative leap. A total "hook, line, and sinker" moment. When that happens, sign me up and tell me where to bring my money.

 

[sfogg]

Thus making the situation of the sweet spot even more fragile - now I need to worry about sitting in the double sweet spot and if I move left or right I ruin both.

No, it doesn't. I say that with 28 years experrience listening to music with a center speaker. Not hypothetical conjecture.

For example, go find reviews talking about Meridian's Trifield. Meridian actually had a little traction getting that setup reviewed for music. Trifield is great in what it does for L/C/R on music. You won't find a single one saying the soundstage is more fragile, quite the opposite. Same for Logic 7 and for the most part DPLII as well.

Take a widely placed pair of speakers and the phantom imaging is more fragile than a narrowly placed pair of speakers. Obviously, the soundstage is bigger with the wide pair. A center gets you the greater solidity of the narrow placement with the larger soundstage of the wider spaced speakers.

This isn't new thinking. Bell Labs wrote about this in the very early days of stereo, Klipsch wrote about this in the 50s.

I am.

Still doesn't fix that L/R phantom imaging gives conflicting HRTF compared to an actual center.

As I said, I don't disagree with adding more physical speakers in principle, I just don't see much payoff - for myself - in going from two in the front to five or so. But going from five to hundreds of individually controlled vertical line sources is a whole different ball game as long as they are spaced tightly enough to avoid diffraction and lobing at the wavelengths that matter. It is an old idea, but still mostly in the research domain.

I think Trifield was actually designed for any number of front speakers but I don't think any commercial product ever implemented that. Kind of surprised nobody has built it DIY (that I know of) as it can be done in the analog domain. Meridian had one analog processor with it.

Your entire front wall covered with narrow vertical speaker strips, maybe a hundred per meter. Each strip is individually driven, so the wall becomes a phased array. The wall can launch flat sound waves from any horizontal angle, combine multiple ones at once, and essentially program the sound field. No more fragile sweet spots, no more comb filtering, no more relying on phantom images holding together - just stable, controllable acoustics. Finally, the actual, real object oriented solution that is what it says it is.

That is why, when this trickles down to consumer scale (even in a cut-down form), it will not be just another incremental upgrade. It will be a qualitative leap. A total "hook, line, and sinker" moment. When that happens, sign me up and tell me where to bring my money.

And yet some people would still be complaining that it doesn't sound like 2 channel.

 

[Newman]

I have come to the conclusion that the OP is argumentative and resistant in relation to the thread topic. We all know where that ends up on the modern internet. It ceases to be a process of inquiry and learning.

At least third party readers will be able to see the points, that he cannot concede, being made by several thread contributors.

 

[slaweks]

How does it help with symphonic recordings? Or with opera, where multiple performers are moving across the stage? A center speaker can lock one source, sure - but in those cases the whole point is a wide and/or shifting soundstage. What exactly is the center "fixing" there?


If you’re building a case for going from 2 to 20 speakers across the front hemisphere, then I don’t disagree in principle

But I’m doubtful that going from 2 to 3 makes a huge difference for classical, choral, or anything with multiple distinct, equally important sources spread left to right - especially when there’s nothing of critical importance parked in the dead center.


Yes, and I've never said otherwise.


Of course I am - the propagation delay I mentioned already accounts for both the spacing between the speakers and the spacing between the ears.

In opera people move, but in oratorios, and other choral works they do not. The center is great to feature soloists.

 

[tengiz]

I have come to the conclusion that the OP is argumentative and resistant in relation to the thread topic. We all know where that ends up on the modern internet. It ceases to be a process of inquiry and learning.

At least third party readers will be able to see the points, that he cannot concede, being made by several thread contributors.

On the opposite, I've learned enough to restrain my initial enthusiasm for Atmos in music, especially for the genres and forms I care about most.

As for conceding, please bring better technical arguments and I will be happy to oblige. But as long as the discussion sits partly in the subjective realm of preference, and as long as technical arguments that aim for a nuanced understanding are brushed aside rather than convincingly dissected, I am not sure what you expect me to do.

 

[tengiz]

No, it doesn't. I say that with 28 years experrience listening to music with a center speaker. Not hypothetical conjecture.
...
Same for Logic 7 and for the most part DPLII as well.

Of course I don't doubt that is your experience. I, on the other hand, very rarely find technologies like Logic 7 or DPLII working well to my taste. It does occasionally happen for me, but usually it has nothing to do with the center channel coming alive in stereo recordings - it is more about what happens in the surrounds.

Still doesn't fix that L/R phantom imaging gives conflicting HRTF compared to an actual center.

The essence of my point is this: using comb filtering as the justification for a physical center speaker is, in my view, lopsided. Comb filtering is not unique to the phantom center - it is present with any phantom source created by more than one speaker, with the keyword here being crosstalk.

The angle simply shifts where the effect sits in frequency and how pronounced it is, but it never disappears. Yes, the real center fixes the phantom center problem. But the moment you use L/C or C/R to create new phantoms, you are right back in the same territory, only now with an extra wrinkle: each of these phantoms is now asymmetrical with respect to the ears, so the comb filter structure itself is different left vs. right.

Yes, I get what you are saying - that despite being flawed in a different way it is still better. Maybe that is true for most. I have not been able to experience it, unfortunately.

And yet some people would still be complaining that it doesn't sound like 2 channel.

Absolutely - I can also imagine the deep disappointment of those not willing to give up their height channels in exchange for an otherwise perfect wall of sound.

 

[tengiz]

In opera people move, but in oratorios, and other choral works they do not. The center is great to feature soloists.

True. But what happens when there are four static soloists with equally important parts?

 

[sfogg]

Of course I don't doubt that is your experience. I, on the other hand, very rarely find technologies like Logic 7 or DPLII working well to my taste. It does occasionally happen for me, but usually it has nothing to do with the center channel coming alive in stereo recordings - it is more about what happens in the surrounds.

What Lexicon did you have? Lots of settings to adjust how it runs the center. You can of course also turn it off and just enjoy the surrounds.

The essence of my point is this: using comb filtering as the justification for a physical center speaker is, in my view, lopsided. Comb filtering is not unique to the phantom center - it is present with any phantom source created by more than one speaker, with the keyword here being crosstalk.

The angle simply shifts where the effect sits in frequency and how pronounced it is, but it never disappears. Yes, the real center fixes the phantom center problem. But the moment you use L/C or C/R to create new phantoms, you are right back in the same territory, only now with an extra wrinkle: each of these phantoms is now asymmetrical with respect to the ears, so the comb filter structure itself is different left vs. right.

I think you are missing this again. Having the phantom be asymmetrical is a good thing. It is closer to how our localization works. If someone is speaking to you off center to the left your left ear hears direct sound and your right ear hears the sound with the HRTF and ITD. We localize it that way.

In stereo for an off center vocal the phantom is of course coming from L/R. Each ear is going to hear direct sound from the speaker on its side and the other ear is going to hear the other speakers sound with the HRTF and the ITD from each speaker. It is conflicting info.

Image between L and C and you are closer to what happens in the real world as you are cutting down the differences in angles again. From the left speaker you hear direct sound in your left ear and the right ear hears it with the HRTF and ITD. The center gives you no ITD and the HRTF is the same for both. It gives less bad info than what happened in stereo.

Yes, I get what you are saying - that despite being flawed in a different way it is still better. Maybe that is true for most. I have not been able to experience it, unfortunately.

Absolutely - I can also imagine the deep disappointment of those not willing to give up their height channels in exchange for an otherwise perfect wall of sound.

I would want surround info too, not just all front. I don't think all front can get me the envelopment that surround can pull out.

 

[tengiz]

I think you are missing this again. Having the phantom be asymmetrical is a good thing. It is closer to how our localization works. If someone is speaking to you off center to the left your left ear hears direct sound and your right ear hears the sound with the HRTF and ITD. We localize it that way.

In stereo for an off center vocal the phantom is of course coming from L/R. Each ear is going to hear direct sound from the speaker on its side and the other ear is going to hear the other speakers sound with the HRTF and the ITD from each speaker. It is conflicting info.

Image between L and C and you are closer to what happens in the real world as you are cutting down the differences in angles again. From the left speaker you hear direct sound in your left ear and the right ear hears it with the HRTF and ITD. The center gives you no ITD and the HRTF is the same for both. It gives less bad info than what happened in stereo.

I don't think I am missing it. I just don't see how this argument addresses what I have said.

The asymmetry between left and right in the case of a single physical source somewhat to the side is a different matter. With one real source, the difference between ears is shaped by the HRTF (which by definition already includes the delay, level differences, and spectral effects), and the brain handles that naturally. With a phantom created by two speakers, you also add crosstalk interference artifacts that result in comb filtering on top of that.

I would want surround info too, not just all front. I don't think all front can get me the envelopment that surround can pull out.

Just add another wall of sound behind

 

[sfogg]

Which Lexicon did you have?

I responded to your comment based on the reality of phantom imaging in L/R vs phantom imaging in L/C/R. Getting back to the original question of the thread.

Totally, agree that the less phantom imaging (or reflections) we need to use to try and reproduce *all* of what is in a recording the better. I literally just added10 additional channels to work toward that along with some very fancy processing.

 

[slaweks]

True. But what happens when there are four static soloists with equally important parts?

If they sing together, then, I guess, each of them can have own microphone, and then you mix it to the center channel. Still, center channel.
I can hear soloists in the center channel and the choir in the main and wides in Schwedischer Rundfunkchor, Berlin Philharmonic & Claudio Abbado - Mozart: Requiem.

 

[tengiz]

If they sing together, then, I guess, each of them can have own microphone, and then you mix it to the center channel. Still, center channel.

That works. However, it would be the opposite of what I (and many others) find most enjoyable in this genre of recordings - the ability to spatially separate the voices. Not only does that sound more natural - unless you are fine with the impression of four people piled on top of each other - it also improves the intelligibility of the polyphony and gives a more realistic perception of vocal harmonies.