I should have explained that better in my post.
The source of the sound was a videogame.
So they basically had cheap/tiny speakers set up in a ring, and since all the sounds were being generated in real time, the feeling of being "in the game" was crazy.
I've been in plenty of movie theaters with Atmos where they had speakers in all directions, but Atmos didn't sound remotely as pinpoint.
I think the key to the whole setup was that the room was absolutely huge (CES auditoriums are the size of an airplane hangar) so there were virtually no reflections to muddy the location of the sounds. It was wild.
I wonder if something comparable could be accomplished in a normal room using cardioids or possibly dipoles?
Another day listening and I can hear better what pushing that 10 up to 3500 is doing. For one thing it's covering up for too much energy from my horns at around 2500 Hz. So I did a little adjustment on the horn EQ and now I'm really digging this setup with the horns all the way down to 1200Hz. The 10s do get a recessed effect when pushed up so high because of the narrow directivity, as expected, and that ends up sounding best at lower volume where it gets that airy effect. At regular listening levels the mids can be noticeably weak but the imaging remains excellent so it's ... interesting. Directivity can be used as a sort of loudness contour. Either way, I think I prefer the crossover up at 3500 or down at 1200, not places in the middle.
Your question is a very valid one.
L+R from the centre interferes with L-R from the left and R-L from the right to make 2L and 2R respectively. But why L-R and R-L won't cancel each other similarly is a mystery !!
To better understand how this works I used an on-line wave simulator to see how the waves interact when something is panned to the side. If a sound is hard panned left you end up with the array playing
L on left,
L on center, and
-L on right.
This creates a null dead center between the left and right channels, but the interesting thing that happens when the center fills in the null is that it makes it narrower and creates a loud zone on the left side of the null, and a quieter zone on the right side of the null. Since the null itself becomes narrow, it's easy for each ear to be in the appropriate zone so that the left ear hears the signal loudly and the right hears it quieter and delayed in phase. This makes for a strong enough stereo effect to create a nice wide sound stage with no unwanted interference nulls at either ear to screw with the tone. If the sound is panned center, then the sound only comes through the center channel so there are no interference patterns at all. So you get smooth tone across the sound stage with a rock solid center.
With regular 2 speaker stereo things work out perfectly when sounds are panned either hard left or hard right. There are no interference patterns and the sound is actually coming from where it's supposed to. When things are panned center, that's when it gets messy. The phantom center imaging has both speakers playing the same sound so they create interference patterns with a lobe dead center. As the frequency goes up the interference pattern nulls get closer together until eventually your ears end up in one and you hear comb filtering.
In both cases you get crosstalk but there's another distinct difference. With regular stereo the first thing you hear is the left signal in the left ear and right signal in the right ear, which is as it should be. But about 0.23 milliseconds later you end up hearing the wrong speaker in each ear as the sound crosses your head. This interferes with imaging as you now have a highly coherent reverse image trying to compete with the initial signal.
With my 3 speaker system the crosstalk is very different. The first thing you hear is the left channel in the left ear and the right channel in the right ear. But then about .2 milliseconds later (maybe less) you hear R-L in the left ear and L-R in the right ear. This means both ears are hearing both channels out of phase across the head, which destroys any location information. So that means the crosstalk that does reach your ears comes across as spacious but without location information, so it doesn't compete with the initial signal that does contain location information.
Popular Mechanics had an article on this in the 70/80 when surround sound home theater was very new and expensive and described this as a poor man’s center channel or ‘null’ channel and when I was young and installing car stereos it was sold to many car stereo customers as 5.1 or surround systems were not yet available for cars. It basically worked about as well as later digital surround systems for cars, which is it say it had little effect in cars. The near field closest that is a car did not see much benefit.
Does anyone have knowledge of the impedance change the amp might see using this configuration?
Does anyone have knowledge of the impedance change the amp might see using this configuration?
I did some work tonight with REW, calculating the predicted combined response at each ear based on the differences in timing of each signal from the 3 speaker array with about 1 foot spacing and listing about 7 feet back to a side panned signal. I think this clearly shows why it works so well. Below 1000 Hz it's mostly phase that gives us directional info. At 1500Hz it's a mix. By 2000 it's mostly level difference. You can see that the array starts to create a level difference around 1000 Hz that gets stronger for a while until it looses power, and eventually at higher frequencies actually goes backwards. This has prompted me to roll off the extreme highs on the side channels. I thought I was hearing that! My attempts at measuring were showing that reversal too.
In any case, I think this calculated response difference of between each ear for a side panned signal provides a solid explanation why the 3 speaker array works to create a wide stereo soundstage. This doesn't show the phase difference that's happening below 1000Hz, but it is happening. I should add that I think if you start getting strong level differences below 1000Hz through recursive crosstalk reduction it's going to create a strong impression of sound sources right next to your ear, because that's normally the only way that can happen. So it's great for special effects, but for normal musical listening we don't need that much crosstalk reduction.
View attachment 1230401
In any case, I think this calculated response difference of between each ear for a side panned signal provides a solid explanation why the 3 speaker array works to create a wide stereo soundstage. This doesn't show the phase difference that's happening below 1000Hz, but it is happening. I should add that I think if you start getting strong level differences below 1000Hz through recursive crosstalk reduction it's going to create a strong impression of sound sources right next to your ear, because that's normally the only way that can happen. So it's great for special effects, but for normal musical listening we don't need that much crosstalk reduction.
View attachment 1230401
If your amplifier is BTL, that would give an individual +L, -L, +R and -R. Now, one would think the amp wouldnt care if you connected a pair (+, -) to +R, -L and +L, -R as a bridge tie for mono. Or +L, +R and -L, -R for the L - R, R - L part. Have I got that right? Would this arrangement work using FR drivers, versus a 2-way? Thanks!
The only ‘problem’ this poses is the system is a one listener affair……move off axis 15 degrees and the whole thing falls apart. I experimented with this for days while I suffered with Covid and basically was isolated with nothing else to do but listen to music and watch movies.
I kept the 3 speaker variant in place and installed a simple Bluetooth enabled remote single pole switch so when I’m listening to the mains for music, I simply disconnect the the center with the remote. It’s my man cave system so there‘s no off axis violators of my personal space.
But I’m currently building a 5 element horizontal Bessel array center using 3” wideband drivers which will be much smaller and I think will have excellent off axis response. Bessel arrays according to Keele need about 10x the width of the array to average out and produce sound like a single point source……we shall see.
I kept the 3 speaker variant in place and installed a simple Bluetooth enabled remote single pole switch so when I’m listening to the mains for music, I simply disconnect the the center with the remote. It’s my man cave system so there‘s no off axis violators of my personal space.
But I’m currently building a 5 element horizontal Bessel array center using 3” wideband drivers which will be much smaller and I think will have excellent off axis response. Bessel arrays according to Keele need about 10x the width of the array to average out and produce sound like a single point source……we shall see.
Something else to play with -pre-dates Hafler / Gerzon. Leonard Feldman came up with it -or at least probably deserves to be remembered for it, insofar as anybody does at all other than sad souls like muggins here (improbable). 
Variation on width stereo. Regular stereo setup at the front. Derive / matrix a centre sum-channel from the L+R, then place that centre channel directly behind you, rather than in front. Don't even go there for speech / movies etc., and as Michael Gerzon said (at least with the ESL-57s and conventional speakers he used in the early '70s), the results can be fine 'only if one clamped one's head into the confines of the one square inch in which it worked'. Feldman however evidently found the results acceptable when he was experimenting, and presumably helped inspire his contribution to EV-4 / Stereo-4 (he was also largely responsible for width stereo with a 2nd independently powered stereo system, at a lower playback level placed behind the listener), so this may have been a question of what speakers & their power / polar responses were used, playback material etc. -Gerzon was notoriously picky, and had little time for most classical recordings coming from the US for e.g. Either way -I remember trying it out a few years ago & while it's 'best' with orchestral, and definitely with independent power, it was interesting & not as bad as Gerzon said in terms of positioning -let alone as bad as matrix quad with no logic decoding could be in terms of channel separation. Since it doesn't cost anything -bit of fun. If nothing else, playing with the various speaker-level matrix setups (I still sometimes use a Hafler / Gerzon difference-signal rear channel) gives you a lot of hands-on insight both into how stereo itself works but also revealing how some engineers were working at the mixing desk. Within reason, I'm a fan.
Variation on width stereo. Regular stereo setup at the front. Derive / matrix a centre sum-channel from the L+R, then place that centre channel directly behind you, rather than in front. Don't even go there for speech / movies etc., and as Michael Gerzon said (at least with the ESL-57s and conventional speakers he used in the early '70s), the results can be fine 'only if one clamped one's head into the confines of the one square inch in which it worked'. Feldman however evidently found the results acceptable when he was experimenting, and presumably helped inspire his contribution to EV-4 / Stereo-4 (he was also largely responsible for width stereo with a 2nd independently powered stereo system, at a lower playback level placed behind the listener), so this may have been a question of what speakers & their power / polar responses were used, playback material etc. -Gerzon was notoriously picky, and had little time for most classical recordings coming from the US for e.g. Either way -I remember trying it out a few years ago & while it's 'best' with orchestral, and definitely with independent power, it was interesting & not as bad as Gerzon said in terms of positioning -let alone as bad as matrix quad with no logic decoding could be in terms of channel separation. Since it doesn't cost anything -bit of fun. If nothing else, playing with the various speaker-level matrix setups (I still sometimes use a Hafler / Gerzon difference-signal rear channel) gives you a lot of hands-on insight both into how stereo itself works but also revealing how some engineers were working at the mixing desk. Within reason, I'm a fan.
I'm glad you tried it. I wasn't sure if anybody had heard it for themselves because I haven't read any feedback until now. Yes, it's a one listener affair. Many hifi systems are just that. I sit off axis when my friends are over so they can have the good seat. I just have to believe they're hearing the stereo sound because it's almost mono when off axis. A standard 3 speaker up-mix with the speakers spread apart a normal amount is definitely better for off axis listening. I find a standard 2 speaker setup pretty bad off axis too, but at least there is some kind of stereo effect. It's skewed in a way that I'd almost rather there wasn't. But, center panned vocals sound better off axis even though they are no longer centered. Noticing that got me interested in mixing them down to mono, which started this whole affair for me.
Were the speakers right in front of the driver, or centered on the dashboard? Three speakers?
Two 4” drivers of low capacity, free air, attached to the center of the dash with something like plumber’s tape.
The main drivers were a pair of 8”/passive radiators in boxes under the hatchback. Driven by a pair of 200w amps.
I don’t recall any time alignment, having been 1989 or so, but the midbass tied in perfectly with the smaller speakers in front.
We would drain the car battery listening to this rare ABC tape, as the presentation was so rewarding.
The main drivers were a pair of 8”/passive radiators in boxes under the hatchback. Driven by a pair of 200w amps.
I don’t recall any time alignment, having been 1989 or so, but the midbass tied in perfectly with the smaller speakers in front.
We would drain the car battery listening to this rare ABC tape, as the presentation was so rewarding.
You inspired me to try something I've been thinking about for a while; adding side ambient channels, delayed and attenuated standard left and right. Tonight I employed a couple of Sony SSCS-5 to try it, with about -6 dB and a 20ms delay compared to the center. It sounds really nice. I've got the little Sonys way out in the corners, up high. The big question was if it helped to give a good stereo effect off axis. Yes it does! There's a much better sense of soundstage off axis while keeping center panned sounds in the center. Even way off to the side of the room I still hear a nice wide effect with the center staying put. Turning off the side ambient channels for comparison I can tell I'm going to be using them. This is just a richer experience with them on, and highlights a problem with the 3 speaker center array not providing any real energy to hit the ears from the sides. It's vital to have that, so what I have here now is a fancy center channel with delayed and attenuated standard stereo side channels - spaced very wide.
I tried just using the L+R alone in the center with the side ambient Sonys and that ain't bad! I like it best with all the speakers running, no matter where I listen, but I think it's worth trying a delay on the side channels if someone just has a L+R center. I'm just loving being able to listen all around the room and hear a good soundstage.
If you ever get the chance, swap out the Side Ambient (cone drivers) speakers with DML panels. Cone drivers for ambience are nice but DML's sparkle, in comparison.
For me, placing the DML's with the flat surface facing the listening position makes the front wall of the room disappear, more 3D like, and it still keeps the centre image intact.
Placing the DML panels with the Edge of the panel facing the listening position starts giving the impression that the side walls of the room have disappeared, at the cost of a less focused centre image. This is nice when there is no Singer in the music that would normally be placed at centre stage.
Using Cone drivers for direct sound and DML's for ambience, and with a separate amp running each, you get to quickly vary the volume/balance between them both.
The DML panels sound interesting. I'll look in to that. Today I experimented with adding two more SSCS-5, using a second surround receiver to play around with surround modes. Whatever I tried, adding two more speakers in that fashion was a considerable downgrade to the sound. I listened to that all day today trying many settings, levels and time delays. When I finally turned off the surrounds it was a blessed relief. I'm not sure if it's just too many speakers starting to throw the imaging off, or if has to do with the surround modes I had to employ to activate the 2nd set of side speakers. I suspect it was the surround processing I didn't like, although it may also have to do with having extra speakers too far off to the sides. I always prefer all the sound sources to be ahead of my listening position.
If you haven't had chance yet, you might want to look at a difference-signal rear channel at a low level (one or two rear speakers), or Gerzon's slightly more advanced version (2 rear speakers only as it's a hybrid of this & width stereo). These tend to be the most consistent / reliable derivations of all from nominal 2-channel, although they're at their best with live recordings, ambient material or material that still has SQ, RM/QS, EV-4 etc. encoding.
That apart, the only other one I've found I can actively like are front-width stereo: regular width stereo but with the 2nd stereo system speakers moved to the front instead of being behind the listener, positioned slightly to the sides of the existing speakers, no processing, but at a slightly lower level. Generally needs identical speakers for best results though. Trad. biamplification (nothing to do with what's currently called biamplification) is a variation on this, but very difficult to try unless you either have suitable rooms, or can do some physical construction work.
That apart, the only other one I've found I can actively like are front-width stereo: regular width stereo but with the 2nd stereo system speakers moved to the front instead of being behind the listener, positioned slightly to the sides of the existing speakers, no processing, but at a slightly lower level. Generally needs identical speakers for best results though. Trad. biamplification (nothing to do with what's currently called biamplification) is a variation on this, but very difficult to try unless you either have suitable rooms, or can do some physical construction work.
I've got everything I need to give it a try except the amplifier channels. I'm thinking about getting a USB to HDMI dongle for my Mac Mini to allow direct access to all 5 channels on the 2nd receiver. I don't know for sure if the USB to HDMI will do that. The close-in doubling of th stereo speakers sounds interesting. I wonder why that works.