The “Real” Speaker Preference Test

[TimVG]

True, but it is not the individual shapes of these curves that convey the key fact, it is the observation that evidence of resonances - the most audible problem in loudspeakers - is mostly revealed in large spatial averages, which all these curves are. If a spike, bump or hump persists from on-axis through listening window and early reflections (closely equivalent to PIR that is dominated by it), to sound power (the ultimate spatial average) it is definitely a resonance. Smoothness and flatness in the direct sound is important (on-axis and listening window), and smoothness in the other larger spatial averages, all of which are indications that the loudspeaker lacks resonances. The actual shapes of the off-axis curves is of value, but definitely secondary, as they relate to later arriving reflected sounds.

So, have a look at Sean's findings and the words smoothness (SM), flatness (AAD) and narrow-band deviation (NBD) appear in front of the important curves. The model is finding resonances, but doesn't use the word.

Here is an example of a loudspeaker that would do badly in the model with the resonances identified. Highly rated loudspeakers exhibit smooth curves, beginning with a flat direct sound.

View attachment 521589

Dr. Toole, I was wondering if you have any thoughts on the behaviour of non cone/dome type loudspeakers and if further preference studies are worth pursuing. In a different thread I raised some questions based on the available research, and how since then, different types of loudspeakers are now more commonly available (cardioid models, to give an example). These loudspeakers tend to produce in-room (predicted and measured) slopes that are more shallow than one would expect out of a traditional loudspeaker. On the other end of the spectrum, I've seen comments about small loudspeakers with "too deep" waveguides that, while also producing smooth curves, no longer adhere to the preferred slopes from the studies performed by Dr. Sean Olive

Here is such an example where the orange and blue shaded slopes are the "ideal" according those studies.

So to be clear, not questioning neutral on-axis/smooth off-axis, but rather how different types of loudspeakers affect the steepness of the off-axis behaviour, and how listerener preference is affected by this, since there are now a lot more neutral loudspeakers commonly available than when those studies were performed.

Thank you.

 

[Floyd Toole]

Dr. Toole, I was wondering if you have any thoughts on the behaviour of non cone/dome type loudspeakers and if further preference studies are worth pursuing. In a different thread I raised some questions based on the available research, and how since then, different types of loudspeakers are now more commonly available (cardioid models, to give an example). These loudspeakers tend to produce in-room (predicted and measured) slopes that are more shallow than one would expect out of a traditional loudspeaker. On the other end of the spectrum, I've seen comments about small loudspeakers with "too deep" waveguides that, while also producing smooth curves, no longer adhere to the preferred slopes from the studies performed by Dr. Sean Olive

Here is such an example where the orange and blue shaded slopes are the "ideal" according those studies.

View attachment 521602



So to be clear, not questioning neutral on-axis/smooth off-axis, but rather how different types of loudspeakers affect the steepness of the off-axis behaviour, and how listerener preference is affected by this, since there are now a lot more neutral loudspeakers commonly available than when those studies were performed.

Thank you.

There is no justification in using these off-axis/directivity slopes as figures of merit unless on puts restrictions on the configuration of the loudspeaker. The curves that Sean was working with, and that I used when I opened Pandora's box by noting that steady-state room curves could be predicted from anechoic data back in 1986 - in spinoramas just the early-reflections curve suffices as a predictor - were derived from what I would now call "last generation" designs, before the almost universal adoption of tweeters with generous waveguides ("horn" sounds so crude in this context ). The waveguides greatly improved the directivity match at crossover, but they also resulted in much more constant directivity above crossover. The result, flattish room curves above crossover.

Oh dear, what do we do now if we put value in the gentle downward slope, because a flat direct sound is desirable, and there is evidence for this?

I have tried to undo the problem I created by stating that room curves are results, not targets, but the horse has left the stable, and people don't read. In addition, tweaking room curves has become a business for some and a hobby for many. "What is your favourite room curve" is a popular forum topic. Opinions formed in normal listening circumstances are "flexible" and listeners adapt, so the myth is perpetuated. Altering a room curve also changes the more important direct sound - maybe not a good idea. All that said, room curves are increasingly definitive information below about 500 Hz.

If the Harman research facility still existed and were operational, it might be possible to test directivity as a somewhat independent variable, but making it a totally independent variable would be a challenge. Tiny changes in broadband spectral balance are easily heard and, in my estimation, would or could nullify significant differences in off-axis directivity. It remains an open question, and in the real world differences in the shapes, sizes, acoustical treatments of rooms, and physical arrangement of the playback system within them might ultimately be the dominant factors. I am willing to guess that minor tweaks of tone controls and/or equalization, combined with the ability of humans to adapt to "innocent" differences, would make the "problem" disappear, so long as the loudspeaker is fundamentally competently designed.

The cardioid speakers I am aware of have that directivity pattern only at low frequencies, where it extends significant directivity to lower frequencies. Combined with a horn-loaded tweeter the DI can be very constant over most of the frequency range. It is an advantage in minimizing the low-frequency boundary effect of the wall behind the loudspeaker. The energy inserted into the room modes is reduced, but since bass EQ is a requirement in small rooms I doubt that it is a consequential advantage. The origin of the bass energy is the Left/Right stereo locations, not where one would choose to place multiple subwoofers to optimally address room resonance problems - see Chapter 14. So-o-o. It is a good idea, costs money, but not a complete solution.

 

[TimVG]

There is no justification in using these off-axis/directivity slopes as figures of merit unless on puts restrictions on the configuration of the loudspeaker. The curves that Sean was working with, and that I used when I opened Pandora's box by noting that steady-state room curves could be predicted from anechoic data back in 1986 - in spinoramas just the early-reflections curve suffices as a predictor - were derived from what I would now call "last generation" designs, before the almost universal adoption of tweeters with generous waveguides ("horn" sounds so crude in this context ). The waveguides greatly improved the directivity match at crossover, but they also resulted in much more constant directivity above crossover. The result, flattish room curves above crossover.

Oh dear, what do we do now if we put value in the gentle downward slope, because a flat direct sound is desirable, and there is evidence for this?

I have tried to undo the problem I created by stating that room curves are results, not targets, but the horse has left the stable, and people don't read. In addition, tweaking room curves has become a business for some and a hobby for many. "What is your favourite room curve" is a popular forum topic. Opinions formed in normal listening circumstances are "flexible" and listeners adapt, so the myth is perpetuated. Altering a room curve also changes the more important direct sound - maybe not a good idea. All that said, room curves are increasingly definitive information below about 500 Hz.

If the Harman research facility still existed and were operational, it might be possible to test directivity as a somewhat independent variable, but making it a totally independent variable would be a challenge. Tiny changes in broadband spectral balance are easily heard and, in my estimation, would or could nullify significant differences in off-axis directivity. It remains an open question, and in the real world differences in the shapes, sizes, acoustical treatments of rooms, and physical arrangement of the playback system within them might ultimately be the dominant factors. I am willing to guess that minor tweaks of tone controls and/or equalization, combined with the ability of humans to adapt to "innocent" differences, would make the "problem" disappear, so long as the loudspeaker is fundamentally competently designed.

The cardioid speakers I am aware of have that directivity pattern only at low frequencies, where it extends significant directivity to lower frequencies. Combined with a horn-loaded tweeter the DI can be very constant over most of the frequency range. It is an advantage in minimizing the low-frequency boundary effect of the wall behind the loudspeaker. The energy inserted into the room modes is reduced, but since bass EQ is a requirement in small rooms I doubt that it is a consequential advantage. The origin of the bass energy is the Left/Right stereo locations, not where one would choose to place multiple subwoofers to optimally address room resonance problems - see Chapter 14. So-o-o. It is a good idea, costs money, but not a complete solution.

Thank you for the extended reply, Dr. Toole - much appreciated. So the Gemini explanation* (in this case for in-wall models) is actually correct, and the preferred slopes (in terms of regression from flat) from the Olive/Welti research are simply so, because the large majority of the loudspeakers tested were forward firing cone/dome loudspeakers.

* In actual practice, the "target" for in-wall speakers is simply linearity. The research (Welti/Olive) suggests that for in-walls, listeners prefer the smoothness of the slope over the steepness of the slope."

 

[Mort]

Was there not some research that indicated a preference for controlled and/or wide directivity? That might be another myth.

 

[Floyd Toole]

Was there not some research that indicated a preference for controlled and/or wide directivity? That might be another myth.

As I said, to control loudspeaker directivity as a truly independent experimental variable is a challenge - not necessarily impossible, but a challenge. There are examples of manufacturers of loudspeakers capable of varied amounts of sound radiated towards the adjacent sidewall, promoting the virtues of such a feature. I have no problem with the concept, and psychoacoustics would indicate that it would be the adjacent sidewall bounce that would have the greatest effect, not overall directivity of the loudspeaker. It would increase the sense of space and many people find it pleasant, but at the same time it would add ambiguity to the individual soundstage images and other people would find that unattractive. There would be, I can imagine, a significant interaction with the musical genre - spacious classics vs. in-your-face rock.

Put differently, it is not the overall Directivity Index that drives spatial perception, it is primarily lateral reflections, and of those the strongest are the ones from the sidewall adjacent to the L and R loudspeakers. It is a subset of a subset of the anechoic data that is needed to begin the discussion.

Chapters 6 and 7 in the 4th edition discuss the topic, but may or may not answer your question in the terms you want .

 

[kimmosto]

Look at the speaker preference ratings and how they are calculated. For a deeper dive read Floyd Toole's book(s)

I'm familiar with both. Sokel already mentioned - and thousands of users already know that one of my applications is able to optimize crossover and driver locations with SM of five, NBD of two, AAD_ON and SL of two individual responses or power averages (but without few mathematical problems/risks in the original NBD and SM). LFX and LFQ can be included too (but without directivity concept limit in the original). So the calculations have been modified to support speaker designing. Not just for calculating single magnitude smoothess rank without any other natural targets.
I have also the book, but it's not specific. It focuses primarily to conventional boxed speakers or cinema/HT, and does not contain much (or any) targets for example to help selecting concepts, locations, dimensions, radiators, whole construction etc. So "Harman's principle" does not say much in practice. It's just a principle - literally. For example speaker designers could need much more, deeper and concept-specific information and suitable targets for their work. Some designers probably have higher standards and more variables which are worth to optimize. Not just the most significant for common consumer i.e. resonances in magnitude responses and their power averages, and directivity features. At least I have higher requirements - partly because I've listened enough speakers which fulfill perfectly every aspect in "Harman's principle", but are still lame P.O.S. Of course, many people like them due to very neutral tonality (and probably brand too), but that does not mean that there would not be possibilities to much higher.

 

[MattHooper]

So it seems we don’t have any scientifically rigourous information available on how the Harman Kardon project of developing loudspeakers through blind testing translated into consumer preferences. (unless I missed it somewhere.)

But informally/anecdotally, a while back I went looking for reviews of the Revel Ultima Salon 2 to see how they were greeted upon release.

And it turned out that rave reviews in Stereophile, The Absolute Sound, Soundstage, Positive Feedback converged on calling the Ultima Salon2 a new state of the art.

And since that also coincides with how highly regarded those speakers are here, I started a thread asking if it was still seen as a state-of-the-art:

Is The Revel Ultima Salon2 STILL the State Of The Art?

This question arose when I recently went down a rabbit hole of reading a bunch of subjective reviews of the Salon2, around when the speaker was introduced. I got to thinking about the Salon2 when I saw a pair for sale (not that I was going to buy them), and I wondered this: technically, we...

www.audiosciencereview.com

As I pointed out in that thread, it’s not just that the informal-use reviews all wrote raves - I found it interesting that the reviews converged very closely in describing the specific characteristics (or lack of character!) that made it state of the art in their listening impressions.

So in a way at least there’s a little bit of anecdotal evidence that the efforts and goals in the development of those speakers carried over to informal, sighted listening impressions.

I found my own impressions of Revel speakers also seemed to conform pretty well to the blind test data. I heard and auditioned the Revel Performa F228Be speakers in two different audio stores. The first encounter didn’t go very well - the sound was somewhat uneven and a bit coarse in the highs. But this was very obviously due to suboptimal set up - they had been pushed up too close to the wall behind them and it was a bad room with one of the speakers being pushed towards a corner right beside a wall/window that was entirely thick glass. So I pretty quickly gave up trying to get any measure of their sound in that place.

However, I got some really good auditions of the speaker in another audio salon in which the Performa’s were set up sensibly.
And what I perceived there was close to
“ flawless,” similar to the way the magazine reviewers we’re flummoxed by trying to find anything wrong with the Salon Ultra.
The speaker just seemed eminently smooth, neutral, well balanced, coherent… just go down the list. Similar to How Kal Rubinson experienced them

So anecdotally it seems like the Revel design data translated pretty well into what I have perceived in informal listening scenarios, as well as to what a number of audio reviewers have perceived as well.

(I did know before auditioning the Performa speakers about the development process and how well they measured, so that in principal could’ve affected my perception. But against that is my first encounter with them in which they did not sound pleasant, where it seemed good speaker design and expectation effects were not enough to overcome the influence of bad set up in that instance).

 

[illusonic]

As I said, to control loudspeaker directivity as a truly independent experimental variable is a challenge - not necessarily impossible, but a challenge.

The Lexicon SL1 seems the best tool for directivity research as it solves the challenge you mention.

"Ulrich said that the narrow radiation control extends to below 80Hz":

SoundSteer Technology from Lexicon

www.lexicon.com

Here is a picture showing Dr. Horbach with the SL1:

Controlling Directivity with a WMTMW

FollgoTT clued me in to the work of D.B. Keele and Ulrich Horbach, who wrote some papers on how to control directivity of a loudspeaker using a combination of driver spacing and FIR filters...

www.diyaudio.com

With regard to research vs. development, I still vote for the former when it comes to SL1.
Below linked presentation documents part of the research that finally led to the SL1.
It describes vertical directivity control and dates back to 2007. From there it's still a big step to the SL1.

https://www.linkwitzlab.com/Horbach-Keele%20Presentation%20Part%202%20V4.pdf

 

[Duke]

As I said, to control loudspeaker directivity as a truly independent experimental variable is a challenge - not necessarily impossible, but a challenge.

I don't have proof for any of the following so consider it all to just be opinion.

For many years I've been working with multidirectional loudspeakers wherein a "main array" is the only contributor to the direct sound and a "secondary array" contributes only to the reflection field. The two arrays can be adjusted independent of one another, but the secondary array moreso than the main array.

I don't think so much in terms of "directivity" as a metric; rather, I think in terms of "what does the direct sound look like, and what does the reflection field look like?" I believe there is utility in being able to manipulate what's happening in the reflection field without affecting the direct sound.

There are examples of manufacturers of loudspeakers capable of varied amounts of sound radiated towards the adjacent sidewall, promoting the virtues of such a feature. I have no problem with the concept, and psychoacoustics would indicate that it would be the adjacent sidewall bounce that would have the greatest effect, not overall directivity of the loudspeaker.

My finding was that the arrival time of the additional reflection-field energy matters more than the arrival direction, something I learned "the hard way":

Emboldened by what I perceived to be the success of my first bipolar speaker (imagine an approximately-constant-directivity version of the Mirage M1), I built an ambitious "twisted bipole" speaker whose secondary array was aimed at that adjacent sidewall. YES the Apparent Source Width ("soundstage width") increased, and spaciousness did increase overall, BUT it was at the noticeable expense of clarity and image precision and soundstage depth. That being said, this configuration DID work well on the long wall of a very wide room, but not on the short wall of a "normal" room, and I wanted something more versatile as far as playback room size and shape.

Subsequently I began using configurations which aimed the secondary array in directions which resulted in significantly longer reflection path lengths. Overall spaciousness including soundstage depth was subjectively improved a bit more, but without image precision or sound quality (in particular clarity) being degraded unless the secondary array was loo loud. BUT, we lost the increase in Apparent Sourch Width, so arguably it's still a trade-off.

The improvement in sound quality was noticeable but not major, perhaps because the main array had good pattern control to begin with. It was the top end that probably improved the most, as the secondary array had been "voiced" to compensate for the main array's inevitable top-octave beaming.

It would increase the sense of space and many people find it pleasant, but at the same time it would add ambiguity to the individual soundstage images and other people would find that unattractive.

Agreed, that's pretty much what I found with the "twisted bipole" configuration strongly illuminating those first same-side-wall reflection zones.

There would be, I can imagine, a significant interaction with the musical genre - spacious classics vs. in-your-face rock.

That hasn't been my experience, nor the experience of my beta-testers. (←Edited for context and clarity: This statement is made with the more evolved configuration of my multidirectional approach in mind, wherein the secondary array has a much longer reflection path than the earlier (and discredited) “twisted bipole” configuration. And I may have misunderstood what @Floyd Toole meant. Anyway let me explain: I asked my beta-testers specifically whether they found any music genres or even particular recordingswhere the additional reflection-field energy was detrimental to either sound quality or spatial quality with the aforementioned mor eevolved configuration, and they all told me that it was never detrimental. The initial set-up matters, but then it seems to be “set it and forget it”, with no tiresome need to optimize settings from one recording to the next.)

Put differently, it is not the overall Directivity Index that drives spatial perception, it is primarily lateral reflections, and of those the strongest are the ones from the sidewall adjacent to the L and R loudspeakers. It is a subset of a subset of the anechoic data that is needed to begin the discussion.

The best results I've gotten have indeed been when the additional reflection energy arrived from 60 degrees left and right of the centerline - just as you predicted! BUT it had to arrive after a sufficient time delay, otherwise the detrimental side-effects were too great. And of course the level of this additional reflection energy relative to the direct sound has to be correctly set.

In my opinion.

 

[Floyd Toole]

The Lexicon SL1 seems the best tool for directivity research as it solves the challenge you mention.

"Ulrich said that the narrow radiation control extends to below 80Hz":

SoundSteer Technology from Lexicon

www.lexicon.com

Here is a picture showing Dr. Horbach with the SL1:

Controlling Directivity with a WMTMW

FollgoTT clued me in to the work of D.B. Keele and Ulrich Horbach, who wrote some papers on how to control directivity of a loudspeaker using a combination of driver spacing and FIR filters...

www.diyaudio.com

With regard to research vs. development, I still vote for the former when it comes to SL1.
Below linked presentation documents part of the research that finally led to the SL1.
It describes vertical directivity control and dates back to 2007. From there it's still a big step to the SL1.

https://www.linkwitzlab.com/Horbach-Keele%20Presentation%20Part%202%20V4.pdf

I absolutely agree that the SL-1 is a successful design - for its purpose - at least what I was able to hear of it 20 years ago. I knew Ulrich and Don Keele well.

As for research into the effect of loudspeaker directivity, if you have read my book, especially the 4th edition, there is much discussion of what is required to convey the perceptions of being in a large space with an orchestra - the concert hall being the definitive reference. The goal is to deliver to a listener's ears the sounds that would have arrived at the real venue - or - to deliver the perceptually most important sounds, because it turns out that a diffuse sound field is not a requirement for our perceptual systems. It is an interesting exercise, leading quickly to two conclusions on my part.

First: binaural recordings reproduced through headphones with head tracking are best. Crosstalk cancelled loudspeakers reproducing binaural can be a viable option, with some setup constraints. (Section 3.3 in the 4th edition)

Second: loudspeaker reproduction in small rooms requires multiple channels - more than two - probably 5 for concert hall envelopment and 7 for movies requiring persuasive flyovers. (Section 7.2 in the 4th edition). Atmos offers a vertical dimension, mostly useful for sound effects in blockbuster movies. I have been in demonstrations where one could walk around much of the listening room without losing the soundstage and the sense of being "there" in a concert hall. Stereo doesn't make the cut.

But stereo is the default standard, so enthusiasts have sought other ways to expand the soundstage and improve envelopment, the most common is multidirectional loudspeakers - bounce sound off the walls. There are omni, dipole and bipolar designs, and simple arrays spraying sounds in multiple directions, including the controversial Bose 901. Some have designed specifically asymmetrical dispersion patterns that combine a more active somewhat omni sound field with steering of the principal lobe to attempt a bit of time-intensity trading to stabilize the soundstage to entertain multiple listeners (Section 4.3.3 in the 4th edition). It was a mixed blessing, but had a following. Compared to the SL-1 it was a very simple array (it was many years ago), but the SL-1 addresses only a single off centre listener, a different objective. I suspect one could do better with modern technology, but time-intensity trading is a problematic concept when dealing with complex musical sounds.

The real-world problem with any multidirectional loudspeaker is that it depends on the room boundaries to deliver the off-axis sounds. All rooms are different, and most are not as spacious as the ones used to demo the true benefits of such designs. Having the ability to vary the directivity is a convenience, albeit an expensive one.

From the psychoacoustic perspective, multichannel audio is the easy winner - even tasteful upmixing is beneficial for many stereo recordings. All that has been historically missing is a recorded repertoire, but it is significant and growing. Fortunately all movies and most music videos have multichannel sound tracks. When I have visited music and sound recording schools the demonstrations have mostly been multichannel - sometimes many, many channels. Interestingly, gaming employs binaural methods and I have been told that this might be where the "action" is. I am not a gamer. The challenge for the marketplace has been the cost of multiple loudspeakers, and for many people sound quality was compromised. Now, with the advent of meaningful measurements and competent designers, timbrally neutral loudspeakers are available in affordable packages. Without neutral loudspeakers multichannel systems function poorly - the loudspeakers draw attention to themselves. Done well, it is impressive.

In the end one puts one's money where the perceptual rewards are. I have had multichannel audio since 1988, starting with the Lexicon CP-1 (Consumer Product-1). from the mind of another brilliant engineer, Dr. David Griesinger, who also, years later created what has been to me the best stereo upmixer I have experienced (Lexicon Logic 7). As I describe in the 3rd edition, at the time I had built the largest "concert hall" I could afford as our living/dining room (7800 cu ft) and energized it with a pair of almost omni Mirage M1 bipole loudspeakers. It was very entertaining, but really needed more channels (Section 7.4.6 in the 3rd edition). My multichannel system in the home theater was then showing promise and was much more versatile. I have not looked back.

So, as has been said, there are "horses for courses". Set your personal target and there are ways to approach it.

 

[MattHooper]

Definitely horses for courses!

My separate two channel speakers occupy the same room as my surround system.
I love both systems for music.

Real world compromises mean that it is easier to achieve an enveloping, immersive quality with my surround system, and easier to achieve more precise, three-dimensional imaging with the two channel system.

So I experiment with speaker and listener positioning as well as acoustics and reflectivity in my room, to try to enhance immersion and spaciousness of the two channel system. For my money I get surprisingly close to my surround system in terms of the feeling sometimes of the room melting away (at least some portion of it).

 

[Floyd Toole]

easier to achieve more precise, three-dimensional imaging with the two channel system.

Yes, the center channel is a real dilemma for multichannel mixers - the timbre of a real loudspeaker does not match the corrupted timbre of stereo phantom images because of the crosstalk. The "solutions" are to not use it at all, leaving the front soundstage in conventional stereo, use all three fronts for the center image thus superimposing a phantom on a real image, or actually use the center channel but with some electronic trickery to get a better blend - this requires some skill. I often get up and put my ear near the center channel to find out what is done. Laziness usually wins in my experience.

The fact that many center channels don't have the quality of the L & R is one of the common arguments against its use. That was bad marketing of multichannel from the get-go - keep costs down to make the sale. In movies the center channel is the most important one, and in music it is the featured artist.

Any thoughts?

 

[illusonic]

I absolutely agree that the SL-1 is a successful design - for its purpose - at least what I was able to hear of it 20 years ago. I knew Ulrich and Don Keele well.

As for research into the effect of loudspeaker directivity, if you have read my book, especially the 4th edition, there is much discussion of what is required to convey the perceptions of being in a large space with an orchestra - the concert hall being the definitive reference. The goal is to deliver to a listener's ears the sounds that would have arrived at the real venue - or - to deliver the perceptually most important sounds, because it turns out that a diffuse sound field is not a requirement for our perceptual systems. It is an interesting exercise, leading quickly to two conclusions on my part.

First: binaural recordings reproduced through headphones with head tracking are best. Crosstalk cancelled loudspeakers reproducing binaural can be a viable option, with some setup constraints. (Section 3.3 in the 4th edition)

Second: loudspeaker reproduction in small rooms requires multiple channels - more than two - probably 5 for concert hall envelopment and 7 for movies requiring persuasive flyovers. (Section 7.2 in the 4th edition). Atmos offers a vertical dimension, mostly useful for sound effects in blockbuster movies. I have been in demonstrations where one could walk around much of the listening room without losing the soundstage and the sense of being "there" in a concert hall. Stereo doesn't make the cut.

But stereo is the default standard, so enthusiasts have sought other ways to expand the soundstage and improve envelopment, the most common is multidirectional loudspeakers - bounce sound off the walls. There are omni, dipole and bipolar designs, and simple arrays spraying sounds in multiple directions, including the controversial Bose 901. Some have designed specifically asymmetrical dispersion patterns that combine a more active somewhat omni sound field with steering of the principal lobe to attempt a bit of time-intensity trading to stabilize the soundstage to entertain multiple listeners (Section 4.3.3 in the 4th edition). It was a mixed blessing, but had a following. Compared to the SL-1 it was a very simple array (it was many years ago), but the SL-1 addresses only a single off centre listener, a different objective. I suspect one could do better with modern technology, but time-intensity trading is a problematic concept when dealing with complex musical sounds.

The real-world problem with any multidirectional loudspeaker is that it depends on the room boundaries to deliver the off-axis sounds. All rooms are different, and most are not as spacious as the ones used to demo the true benefits of such designs. Having the ability to vary the directivity is a convenience, albeit an expensive one.

From the psychoacoustic perspective, multichannel audio is the easy winner - even tasteful upmixing is beneficial for many stereo recordings. All that has been historically missing is a recorded repertoire, but it is significant and growing. Fortunately all movies and most music videos have multichannel sound tracks. When I have visited music and sound recording schools the demonstrations have mostly been multichannel - sometimes many, many channels. Interestingly, gaming employs binaural methods and I have been told that this might be where the "action" is. I am not a gamer. The challenge for the marketplace has been the cost of multiple loudspeakers, and for many people sound quality was compromised. Now, with the advent of meaningful measurements and competent designers, timbrally neutral loudspeakers are available in affordable packages. Without neutral loudspeakers multichannel systems function poorly - the loudspeakers draw attention to themselves. Done well, it is impressive.

In the end one puts one's money where the perceptual rewards are. I have had multichannel audio since 1988, starting with the Lexicon CP-1 (Consumer Product-1). from the mind of another brilliant engineer, Dr. David Griesinger, who also, years later created what has been to me the best stereo upmixer I have experienced (Lexicon Logic 7). As I describe in the 3rd edition, at the time I had built the largest "concert hall" I could afford as our living/dining room (7800 cu ft) and energized it with a pair of almost omni Mirage M1 bipole loudspeakers. It was very entertaining, but really needed more channels (Section 7.4.6 in the 3rd edition). My multichannel system in the home theater was then showing promise and was much more versatile. I have not looked back.

So, as has been said, there are "horses for courses". Set your personal target and there are ways to approach it.

Thank you for your detailed response. While I have read your book some years ago, 4th edition indeed seems to cover some interesting new topics.

Technologically, SL1 is impressive and from that perspective a successful design. Apparently, Harman did not even start selling it - so no market success. An SL1 demo at the time of product announcement left me in doubt whether the beam steering feature would actually find much use at home. On the other hand, I thought that it could be a useful tool for speaker designers to compare different radiation patterns and come up with fixed radiation targets. Instant switching between controlled radiation patterns for direct comparison is a rare feature. And while smoothly widening or constant directivity are probably most useful, it would certainly be interesting to also examine less perfect radiation shapes as well as di-/bipole and multi-beam patterns based on a loudspeaker array like SL1.

For me binaural HP playback with head tracking usually stayed relatively close to the head (<1m) and frontal image was mostly unstable, if based on foreign HRTFs and artificial reverb. Ambisonics recordings from a spherical mic array worked well despite foreign HRTF (Kemar) used for playback. Natural reverberation seems to help. A Smith Realizer demo worked perfectly to emulate a 7.2.x surround setup. They measured my binaural room impulse responses (BRIR) for every speaker at 3 head angles and compensated the HP response based on an in-ear measurement. Afterwards BRIRs were applied with head tracking. Fast switching between loudspeaker and HP playback allowed direct comparison and the result was almost perfectly identical. Having the speakers as visual cues and the room matching the room response probably helped HP playback. As you mentioned gaming, JBL 3D gaming headphones did not impress me at all. This came as quite a surprise since the Ossic x HP had generated a big crowdfunding hype before Harman hired the key players behind that campaign.

Full envelopment in a room probably requires rear sources as even a multi-beam frontal source can in my experience only cover +/-90° at best. I would expect similar limitations for crosstalk cancellation and binaural beam systems as these will probably have a hard time to reproduce rear sources when playing from the front. It’s in my experience not possible to reliably overcome the personal frontal pinna transfer function attempting to generate rear sources and vice versa. Precise localization, as opposed to mere envelopment, requires physical sources at the respective positions of specific surround formats.

Regarding room-dependency of multi-directional speakers, it may be possible to calibrate beam directions and delays such that reflections arrive at the same time from useful directions. I believe Yamaha did a soundbar with such a feature. Obviously, lots of uncertainties remain.

A general issue with multichannel playback, is the required number of loudspeaker boxes and/or drivers and corresponding electronics. Within a limited budget this will limit quality of components in multi-beam speakers as well as discrete speaker boxes. Planned SL1 price compared to e.g. Revel Ultima Salon2 illustrates this well although SL1 is not even multi-channel. When I listened to SL1 there were Salon2 in the same room. I thought I preferred smooth treble from the beryllium tweeter although that may have been biased perception.

Personally, I still enjoy stereo playback over separate speakers and HP playback with HRTF-based cross feed for some externalization without head tracking. But it’s fun to occasionally experience other options.

 

[Duke]

Yes, the center channel is a real dilemma for multichannel mixers... The "solutions" are to not use it at all, leaving the front soundstage in conventional stereo, use all three fronts for the center image thus superimposing a phantom on a real image, or actually use the center channel but with some electronic trickery to get a better blend - this requires some skill...

The fact that many center channels don't have the quality of the L & R is one of the common arguments against its use...

Any thoughts?

Maybe time-intensity trading, with the processor in "phantom center" mode for movies and music videos?

Assuming suitable loudspeakers, ime this approach offers adequate clarity for movie dialogue and gives acceptably on-screen dialogue location for viewers who do not have a hearing imbalance, even for seats well off the centerline. The benefits include zero sound quality mis-match; no need to figure out where to put the center channel speaker; and (imo) better soundstage depth, which is not a big deal for movies but arguably adds to the enjoyment of music videos.

I think an example of a time-intensity-trading-friendly loudspeaker would be the JBL 4367, and probably some of the smaller models with similar radiation patterns.

 

[MattHooper]

Yes, the center channel is a real dilemma for multichannel mixers - the timbre of a real loudspeaker does not match the corrupted timbre of stereo phantom images because of the crosstalk. The "solutions" are to not use it at all, leaving the front soundstage in conventional stereo, use all three fronts for the center image thus superimposing a phantom on a real image, or actually use the center channel but with some electronic trickery to get a better blend - this requires some skill. I often get up and put my ear near the center channel to find out what is done. Laziness usually wins in my experience.

The fact that many center channels don't have the quality of the L & R is one of the common arguments against its use. That was bad marketing of multichannel from the get-go - keep costs down to make the sale. In movies the center channel is the most important one, and in music it is the featured artist.

Any thoughts?

Nothing very enlightening from me on this.
While I myself do sound design and attend mixes and playbacks in very high-end Studios, I leave the LCR panning/steering to the mixers. As you know, sounds that need to “stick” to the image (eg lots of foley, dialogue, some sound effects) are steered to the centre and hence typically centre channel.

As a home theater enthusiast, I’m happy to have a high quality centre made by a competent company (now defunct: Hales) that is a really excellent “timber match” to their L/R speakers that I also use.
(Photos of large centre speaker here - images are not of mine because I have mine covered in black velvet making it hard to photograph)

When I was designing my home theatre set up I decided to experiment with a wide variety of different L/C/R speakers - various floor standing and stand mount speakers, different dispersion characteristics, even tried Omnis for L/R.

There were certainly some interesting effects and combinations, but “timbre matching” (for lack of better word) was always a tough nut to crack and never quite right when trying to combine different types of speakers.

In the end by far the most satisfying solution were the Hales speakers for LCR. As we know, the mere fact that LCR speakers might be produced by the same company is no guarantee of a seamless match. Often let down by the centre channel design. But I’ve found it can work pretty well in the hands of a competent speaker designer. (I’ve never heard a Revel LCR set up, but I would expect that to be well designed).

I’ve played around sometimes listening to music on my home theatre surround system or movies without the centre channel and just going stereo with the LR. But I inevitably prefer the system with the centre channel.
It not only sounds bigger, but it sounds richer, more timbrally satisfying and solid, the benefit I presume of adding another speaker producing the sound rather than leaving it to the phasey stereo-only performance. (Actually one of my favourite up-mixing modes for stereo on my Denon AVR is “7 Channel Stereo” employing the centre channel as well - sounds huge, rich, more solid than actual stereo with just the L/Rs).

(some people apparently prefer just going stereo in their system for Home theatre duty.
For me that actually worked for a number of years when I started off with a plasma TV and small Spendor 3/5s on either side. Those Spendors did an amazing job of disappearing and somehow mapping the image directly onto the screen, staying convincing even off access. But the much wider spread demanded by my home theatre screen wouldn’t work that way.)

 

[olieb]

It would increase the sense of space and many people find it pleasant, but at the same time it would add ambiguity to the individual soundstage images and other people would find that unattractive.

Agreed, that's pretty much what I found with the "twisted bipole" configuration strongly illuminating those first same-side-wall reflection zones.

There would be, I can imagine, a significant interaction with the musical genre - spacious classics vs. in-your-face rock.

That hasn't been my experience, nor the experience of my beta-testers.

@Duke :
Can you elaborate a bit more on this?
Do you mean that the impression of "more sense of space" versus reduced clarity were independent of genre?
Or does that mean that some people were in favour of same-side-wall reflection and others were not? And this being more a general preference of the listener, not changing with genre?

@Floyd Toole :
And which of the genres would presumably profit (more) from those lateral reflections? Spacious classics or in-your-face rock?

 

[Floyd Toole]

Having the speakers as visual cues and the room matching the room response probably helped HP playback.

Absolutely! Many years ago I learned that even crude binaural recordings made and reproduced from the same head location in the same room were remarkably persuasive, with externalization - until one rotated the head. I discuss it in the 4th edition. Even Jens Blauert got caught by this effect: i.e. everything worked wonderfully until the playback location was changed and the visuals and ambient sounds no longer matched. The brain is a wonderful instrument.

 

[respice finem]

Absolutely! Many years ago I learned that even crude binaural recordings made and reproduced from the same head location in the same room were remarkably persuasive, with externalization - until one rotated the head. I discuss it in the 4th edition. Even Jens Blauert got caught by this effect: i.e. everything worked wonderfully until the playback location was changed and the visuals and ambient sounds no longer matched. The brain is a wonderful instrument.

OTOH, some listeners "early conditioned" for headphones (like myself) may regard it as a feature, to be in any spot in the room without the soundstage changing. At least when only listening to music, without visual content.

 

[Floyd Toole]

OTOH, some listeners "early conditioned" for headphones (like myself) may regard it as a feature, to be in any spot in the room without the soundstage changing. At least when only listening to music, without visual content.

True, but I was focused on the sound sources being perceived as external - " out there" - not inside or close to the head. Interesting that with headphone binaural it is difficult to establish a realistic frontal soundstage - things default to inside or behind, which is why so many binaural demos use clipping hair at the back of the neck. With cross-talk cancelled loudspeakers it is the rear half of the soundstage that is challenging. What is the common factor? Vision: when the eyes see or sense a plausible source of sound the directional perception follows, in the absence of that the survival instinct sets in and perceived localization goes to the rear, in real life causing the head to rotate and resolve the ambiguity - if the sound continues. An isolated twig snap obviously was a concern for our hunter-gatherer ancestors.

And, I agree that headphone listening to stereo source material is an acquired taste - I don't dislike it at all, but it is not stereo, nor binaural, it is what it is.

 

[illusonic]

Absolutely! Many years ago I learned that even crude binaural recordings made and reproduced from the same head location in the same room were remarkably persuasive, with externalization - until one rotated the head. I discuss it in the 4th edition. Even Jens Blauert got caught by this effect: i.e. everything worked wonderfully until the playback location was changed and the visuals and ambient sounds no longer matched. The brain is a wonderful instrument.

I discussed this point with one of the Smyth brothers and he confirmed the issue for some of their customers. They suggest use in a dimly lit room. He also said that it works for him independent of the room and visuals. That’s an effect of the brain learning to rely more on (inter-)aural than on visual cues and also the brain learning a specific set of aural cues. The learning effect can be an issue for researchers in the field. When someone optimizes spatial effects (e.g. generalized binaural processing or simulated room reverberation) while listening over long times, results may work well for that person but much less for others. It’s important to consult others frequently to avoid this trap.