Deep Stereo

[juha_tp]

Hello!

I bumped to this article "A new way to listen to music" by Audioreview (Italy) which introduces an 'effec' they call as Deep Stereo. Original article is written in italian but here's the link for translated (Google) version.

Patent mentioned in article - http://ieeemilestones.ethw.org/images/5 ... OCR%29.pdf

I don't fully understand thepapers to implement it. Has anyone here implemented this already using FS (or SM) ?

[Spogg]

Hi juha_tp

I skimmed the translated version (the patent PDF was far too big and ugly to read ).

I get the idea that this topic might be relevant:

viewtopic.php?f=4&t=8934&hilit=ambiphonics#p35113

Cheers

Spogg

[HughBanton]

That's really interesting. I was particularly fascinated to discover that, way back (even before my time), an attempt had been made to put FOUR channels onto vinyl discs, using the regular FM radio-carrier principle to put two channels on each groove wall. That's gonna work innit ... (4 channels on FM radio would have been even more of a challenge ...)

But it turns out it was the, more natural, 2-channel capabilities of vinyl that has ultimately landed us with the "stereo" we know, meaning '2 speakers' .. whereas the original investigators had been more keen on a 3-channel system for cinema. Who knew. It probably works better.

Converting two channels to three needs the (common) centre information isolating from the left and right channels, putting it on its own channel (C) and then removing it from L and from R.

Do we know how to do that? (I don't!!)

[juha_tp]

...
Converting two channels to three needs the (common) centre information isolating from the left and right channels, putting it on its own channel (C) and then removing it from L and from R.

Do we know how to do that? (I don't!!)

I found one try out as like this:

I am trying to implement Deep stereo concept. There are three speakers instead of two ones. Center speaker should reproduce common sound part, that presents both in Left and Right channels:

DIF_R=-1.0*L+R
DIF_L=L+-1.0*R
LR_CORE=L+R+-1.0*DIF_L+-1.0*DIF_R
L=0.5*LR_CORE R=0.5*LR_CORE

This works. Stereo effect disappeares.

But when I try to use DIF_L and DIF_R for Left and Right channels respectively, they are the same! I use headphones to check the sound:

DIF_R=R+-1.0*L
DIF_L=L+-1.0*R
L=DIF_L R=DIF_R

[HughBanton]

Correct me if I'm wrong but I think there's an obvious falicy here :

LR_CORE = L+R + -1.0*DIF_L+-1.0*DIF_R ???

now the second half says
-1.0*DIF_L+-1.0*DIF_R
= -1.0 *(L+-1.0*R) + -1.0 *(-1.0*L+R)
= -1.0L + R + L + -1.0R
= 0!

therefore
LR_CORE = L+R = mono.

As far as I can see this can't be achieved by any simple matrix method.

[k brown]

Believe it or not, this (interaural cross-talk elimination) can be done almost to perfection with no DSP at all! I discovered this accidentally back in the late '80s when messing around with the old 'Hafler Quad' notion from the early days of stereo. This entailed wiring a center speaker between the ground of the amplifier and the ground connections of the left and right speakers; the 'cold' leads of the L/R speakers are connected together, then to the 'hot' connection of the center speaker; the 'cold' term of the center spkr is connected to amp ground (this whole thing requires an amp that has common-ground speaker connections [most do; not all]. This isolates the L/R speakers from the amplifier with the DCR of the center speaker's voice coil, thus reducing the level of signals that are exactly of equal amplitude in both stereo channels by approx 6dB (somewhat frequency-dependent). Then a fourth speaker is connected across the amp's two '+' terminals, with no connection to ground; this results in near-zero amplitude of centered signals in this speaker, which was placed directly behind the listener. The left and right speakers were placed 90 degrees left of center and 90 degrees right of center; the center, of course directly in front; rear spkr directly behind - thus, a 'diamond' configuration. This creates a generally immersive sound that's quite impressive. BUT --

Something put it in my head to try modifying this arrangement by placing the L/R speakers in a dual version of the usually recommended formation of regular stereo speakers being at a 60 degree angle to the listener. So I tried placing the L and C speakers 60 degrees apart and the C and R speakers 60 degrees apart, forming a 120 arc bisected by the center speaker. All three placed exactly equidistant from the listener for equal time-of-arrival from all three. Well, this was a horse of a different color!! Three-dimensionality beyond belief. Made even better with the rear speaker - also equidistant or slightly farther away. Even better if two speakers used in back, placed about 60 degrees apart and wired with each of their hot terms to their respective amp terms and their cold terms together, usually through a resistor that leaks a small bit ground to them; this results in the left rear reproducing a L-R signal and the right rear a R-L signal. This is starting to sound a lot like the standard 5.1 home theater surround configuration doesn't it? But this was in '88/'89.

It wasn't till a few years later that I learned about binaural, transaural and cross-talk cancellation, and I wondered if this might be what was going on with this setup. I tested it by making sure all the speakers were exactly equidistant then connecting a test signal to the left amp channel only. The result defied belief - it was as if I'd actually lost hearing in my right ear! Not just 'sound from the left of me', but 'I can't hear out of my right ear' - an almost physically uncomfortable sensation. Same for a signal in the right amp channel only. What this meant is that this setup was creating 'electro-acoustic' cross-talk cancellation - anything that appeared ONLY in the left recorded channel was being heard ONLY by the left ear, and vice versa for the right; same as if one was wearing headphones. This particular wide placement of the left and right speakers created just the right time delay (acoustically) for the anti-phase components in each channel to produce the cross-talk cancellation (the 90 degree Hafler L/R placement resulted in too much delay for this cancellation to occur). The rear speakers serve to 'anchor' the phantom sources between the left front and left rear speakers (and vice versa for the right). Any speaker system that produces this 'each ear hears only what's recorded in it's respective stereo channel' is referred to as a Transaural system. The BIG deal about such a system is that the super 3-dimentional illusion from binaural dummy-head recordings that previously could only be experienced with headphones (requires that each ear hear only it's respective recorded channel) can now be achieved over loudspeakers. The most perfect result I've heard was with a system that used three matched front speakers utilising small-diameter full range drivers (so no time-smearing cross-overs) mounted in spherical enclosures (so no time-smearing baffle-edge diffractions). A brilliant recording like Alan Parson's mix on Dark Side of the Moon, was astounding on this - sounded like a high-tech multichannel recording with speakers all around the room, but just a stereo CD and stereo amp (not just a pleasant sense of immersion, but startlingly specific source-placement around the listener). The second-best version of this I had for a while was slightly less perfect cross-talk cancellation wise, but had much wider frequency and dynamic range, was three two-way spkrs with very HQ 4" bass-mids and 1" soft dome tweeters in a biamped system with the three tweeters connected to their own stereo amp and likewise for the bass-mids. This was just a few years before home surround sound systems became common. I called the thing 'Picture Window'.

A rear-speaker variation that almost seemed to work better, was a single open baffle speaker mounted on an approx 12" square baffle (low bass is not required from the rear spkrs) placed on edge to the listener, and a bit above. Since the sound emanating from the front of an open baffle spkr is antiphase to that from the rear, this one driver produces L-R out the left side and R-L from the right.

[wlangfor@uoguelph.ca]

The ramifications of this technique would seem to result in less phase theoretically, due to the fact that the sides are in parallel with the center. That's an interesting idea.

Believe it or not, this (interaural cross-talk elimination) can be done almost to perfection with no DSP at all! I discovered this accidentally back in the late '80s when messing around with the old 'Hafler Quad' notion from the early days of stereo. This entailed wiring a center speaker between the ground of the amplifier and the ground connections of the left and right speakers; the 'cold' leads of the L/R speakers are connected together, then to the 'hot' connection of the center speaker; the 'cold' term of the center spkr is connected to amp ground (this whole thing requires an amp that has common-ground speaker connections [most do; not all]. This isolates the L/R speakers from the amplifier with the DCR of the center speaker's voice coil, thus reducing the level of signals that are exactly of equal amplitude in both stereo channels by approx 6dB (somewhat frequency-dependent). Then a fourth speaker is connected across the amp's two '+' terminals, with no connection to ground; this results in near-zero amplitude of centered signals in this speaker, which was placed directly behind the listener. The left and right speakers were placed 90 degrees left of center and 90 degrees right of center; the center, of course directly in front; rear spkr directly behind - thus, a 'diamond' configuration. This creates a generally immersive sound that's quite impressive. BUT --

Something put it in my head to try modifying this arrangement by placing the L/R speakers in a dual version of the usually recommended formation of regular stereo speakers being at a 60 degree angle to the listener. So I tried placing the L and C speakers 60 degrees apart and the C and R speakers 60 degrees apart, forming a 120 arc bisected by the center speaker. All three placed exactly equidistant from the listener for equal time-of-arrival from all three. Well, this was a horse of a different color!! Three-dimensionality beyond belief. Made even better with the rear speaker - also equidistant or slightly farther away. Even better if two speakers used in back, placed about 60 degrees apart and wired with each of their hot terms to their respective amp terms and their cold terms together, usually through a resistor that leaks a small bit ground to them; this results in the left rear reproducing a L-R signal and the right rear a R-L signal. This is starting to sound a lot like the standard 5.1 home theater surround configuration doesn't it? But this was in '88/'89.

It wasn't till a few years later that I learned about binaural, transaural and cross-talk cancellation, and I wondered if this might be what was going on with this setup. I tested it by making sure all the speakers were exactly equidistant then connecting a test signal to the left amp channel only. The result defied belief - it was as if I'd actually lost hearing in my right ear! Not just 'sound from the left of me', but 'I can't hear out of my right ear' - an almost physically uncomfortable sensation. Same for a signal in the right amp channel only. What this meant is that this setup was creating 'electro-acoustic' cross-talk cancellation - anything that appeared ONLY in the left recorded channel was being heard ONLY by the left ear, and vice versa for the right; same as if one was wearing headphones. This particular wide placement of the left and right speakers created just the right time delay (acoustically) for the anti-phase components in each channel to produce the cross-talk cancellation (the 90 degree Hafler L/R placement resulted in too much delay for this cancellation to occur). The rear speakers serve to 'anchor' the phantom sources between the left front and left rear speakers (and vice versa for the right). Any speaker system that produces this 'each ear hears only what's recorded in it's respective stereo channel' is referred to as a Transaural system. The BIG deal about such a system is that the super 3-dimentional illusion from binaural dummy-head recordings that previously could only be experienced with headphones (requires that each ear hear only it's respective recorded channel) can now be achieved over loudspeakers. The most perfect result I've heard was with a system that used three matched front speakers utilising small-diameter full range drivers (so no time-smearing cross-overs) mounted in spherical enclosures (so no time-smearing baffle-edge diffractions). A brilliant recording like Alan Parson's mix on Dark Side of the Moon, was astounding on this - sounded like a high-tech multichannel recording with speakers all around the room, but just a stereo CD and stereo amp (not just a pleasant sense of immersion, but startlingly specific source-placement around the listener). The second-best version of this I had for a while was slightly less perfect cross-talk cancellation wise, but had much wider frequency and dynamic range, was three two-way spkrs with very HQ 4" bass-mids and 1" soft dome tweeters in a biamped system with the three tweeters connected to their own stereo amp and likewise for the bass-mids. This was just a few years before home surround sound systems became common. I called the thing 'Picture Window'.

A rear-speaker variation that almost seemed to work better, was a single open baffle speaker mounted on an approx 12" square baffle (low bass is not required from the rear spkrs) placed on edge to the listener, and a bit above. Since the sound emanating from the front of an open baffle spkr is antiphase to that from the rear, this one driver produces L-R out the left side and R-L from the right.

[trogluddite]

@k brown
Thankyou for that post. I toyed with the old rear-speaker between the positive poles many years ago, but never experimented further than that. Nor was I aware of transaural reproduction, about which Google is now satiating my curiosity (the first article I found also make me giggle with mention that the method it describes removes "the limitation ... on the number of listeners' ears at two each" - maybe a system suitable for some of Spoggs extra-terrestrial friends? )

One question. I've listened to binaural recording on headphones and find them extremely immersive. That is; until I move my head, which I find extremely disorientating as my brain expects the sound field to keep its "real world" orientation (familiarity may play a part, but I don't recall ever experiencing this with non-binaural recordings). How do you find transaural reproduction through speakers fares in this respect?

[k brown]

Speaker Transaural is actually more restrictive (nothing comes free ); if you move your head the cross-talk cancellation is destroyed and you're back to 'nice sound of speakers all around'. At least with headphones the illusion of what the dummy head recorded remains - it just doesn't change to reflect your changed head position. That's the biggest downside of this - you have sit in a small sweet spot and NOT MOVE. Personally I've never been bothered by it because what I'm hearing is so captivating I don't want to move. I've always been one of those motionless, intensive listeners. I actually had essentially an 'out-of-body' experience from this once. I got so immersed in listening to the Apocalypse Now soundtrack (an unusually three-dimensional recording for a film sountrack (http://designingsound.org/2009/10/07/wa ... orldizing/) that I became convinced that my body was in an entirely different position than it really was. Freaked me out, big time. This was when I was in film school, and my roomates/classmates so identified me with this 'instensive listening' habit that they spoofed it in a film project by shooting me listening motionless to music, then dissolving to a shot of me covered in fake cobwebs.

Even though the actual Transaural/crosstalk cancellation effect requires a very specific listening position, this system is still superior to standard two-speaker stereo when listening off axis. With a regular stereo setup, when you move off the center line, the soundstage quickly collapses to the speaker nearest you; sounds like all the music's coming from that one speaker. With the three-spkr setup this occurs to a far lesser extent.

[Spogg]

This is not entirely on-topic but is a bit related:

When I was little more than a child I saved up and bought two very cheap 12 inch speakers out of Practical Wireless magazine. I thought the bass would be great but it wasn’t, because the magnets were small and the suspension was far too stiff. But what I did find was great stereo imaging.
I eventually realised this was down to the fact that they had concentric tweeter cones; small flared cones glued directly to the voice coil. The treble was surprisingly smooth and accurate too, as far as I could tell.

Later I compared these with new speakers I made which had “proper” tweeters with a passive crossover, and the reproduction was better but the stereo imaging (in the sweet spot) was very inferior. I concluded that frequency-based and spacial-based phase shift is very detrimental to stereo imaging.

Of course I eventually came to know that this is regular knowledge, but it was rather pleasing to have found this out on my own. It leads me to wonder why the full-range single driver concept doesn’t seem to have been developed and adopted.

Cheers

Spogg