Why evaluating the sound of a single speaker is essential

[Floyd Toole]

It's acceptable, but it is so much less common than the myth ("we all hear differently") would have us believe. Indeed, it is fair to call it rare, among people who don't have substantial hearing damage.

Controlled listening has shown that system-wide neutrality is preferred to non-neutrality, and it is common-sense logical too, since we have the real world of natural sound sources as our baseline.

The only way that non-neutral speakers would be preferred would be if some other part of the chain was non-neutral, and the speakers were helping to counter that. That's why I wrote 'system-wide neutrality' above. Start with a neutrally-balanced mastering, recording/playback source, and amplifier, (all of which is common today, especially digitally), and neutral speakers are the only safe bet.

cheers

[edit: I see @Floyd Toole has gazumped me! Read him instead, of course!]

I would add only that the problem with non-neutral loudspeakers is that they add the same monotonous coloration to absolutely everything that is played through them. One can adapt to this to some extent, but in my experience, when something more neutral is heard it is obvious, and preferred. Most people never get to experience a valid, blind, equal loudness comparison test. As Newman said, hearing loss is a factor in preference, but it shows up not only as a fixed bias in preference but also as rating inconsistencies in repeated auditions of exactly the same sound - love it in the morning, hate it in the afternoon has been observed.

 

[Old_School_Brad]

It's acceptable, but it is so much less common than the myth ("we all hear differently") would have us believe. Indeed, it is fair to call it rare, among people who don't have substantial hearing damage.

Controlled listening has shown that system-wide neutrality is preferred to non-neutrality, and it is common-sense logical too, since we have the real world of natural sound sources as our baseline.

The only way that non-neutral speakers would be preferred would be if some other part of the chain was non-neutral, and the speakers were helping to counter that. That's why I wrote 'system-wide neutrality' above. Start with a neutrally-balanced mastering, recording/playback source, and amplifier, (all of which is common today, especially digitally), and neutral speakers are the only safe bet.

cheers

[edit: I see @Floyd Toole has gazumped me! Read him instead, of course!]

Sure, I agree.

My main point was rather to highlight that I haven’t seen anyone here outright criticize someone for liking B&W, Dali, or Cerwin Vega speakers. It’s generally accepted when someone enjoys the sound of a non-neutral speaker.

 

[amirm]

People can absolutely hear tonality differences in stereo pair. This statement is an over exaggeration at best.

They *think* they can when there is no way to verify anything they say. Fortunately we have controlled testing that shows how much the tonal discrimination shrinks as more channels are added:

As you see, by the time you get to multi-channel, it is all gone (everything sounds good). Stereo allows some level but nothing remotely like mono (Situation B).

Add to that the fact that non-trailed listeners do poorly in these tests to begin with, and you can safely surmise you are reading random assessments:

 

[amirm]

We all know the Maggie's are unique design and most certainly require both speakers to possibly enjoy.

Numerous controlled tests by Harman included Martin Logan speakers which are also "unique." They did poorly due to significant colorations/resonances. That, you can't get rid of just because you add another one to the setup:

Speaker "M" is Martin Logan. This is a shot of it when I attended Harman's Blind Listening test setup:

I ranked it extremely low. It almost sounded broken to me.

The thing Maggies have is the false spatial qualities due to back radiation. It sounds super natural on some content but becomes tiresome as it adds the same effect to *every* track you listen to. It is very odd sounding to me when I hear studio produced rock/pop with that kind of effect instead of tight presentation.

 

[Vuki]

They *think* they can when there is no way to verify anything they say. Fortunately we have controlled testing that shows how much the tonal discrimination shrinks as more channels are added:

As you see, by the time you get to multi-channel, it is all gone (everything sounds good). Stereo allows some level but nothing remotely like mono (Situation B).

Add to that the fact that non-trailed listeners do poorly in these tests to begin with, and you can safely surmise you are reading random assessments:

Maybe so, but the first graph also shows that for eq B stereo is much more preffered than mono.

 

[Blumlein 88]

Numerous controlled tests by Harman included Martin Logan speakers which are also "unique." They did poorly due to significant colorations/resonances. That, you can't get rid of just because you add another one to the setup:

Speaker "M" is Martin Logan. This is a shot of it when I attended Harman's Blind Listening test setup:

I ranked it extremely low. It almost sounded broken to me.

The thing Maggies have is the false spatial qualities due to back radiation. It sounds super natural on some content but becomes tiresome as it adds the same effect to *every* track you listen to. It is very odd sounding to me when I hear studio produced rock/pop with that kind of effect instead of tight presentation.

Obviously the result of the blue straps interfering with the sound. Plus you can tell it isn't even aligned evenly with the front edge of the shuffle board it is strapped onto.

Wished they had the MCH results using the M-L (which is not one of my favored ESLs except in CLS version).

 

[amirm]

Maybe so, but the first graph also shows that for eq B stereo is much more preffered than mono.

That's because it hid the colorations that EQ presented. People are blinded by the spatial effects of stereo, not realizing the flaws in the system. Since you can't force content to always present itself this way, then it is a bad way to evaluate things.

But let's go with your opinion. Exactly how do I set up two Maggies as to represent how everyone out there sets them up in stereo???

 

[Vuki]

That's because it hid the colorations that EQ presented. People are blinded by the spatial effects of stereo, not realizing the flaws in the system. Since you can't force content to always present itself this way, then it is a bad way to evaluate things.

But let's go with your opinion. Exactly how do I set up two Maggies as to represent how everyone out there sets them up in stereo???

It just shows that there is at least one loudspeaker significantly preffered listened in stereo than in mono.
How do you set up any loudspeakers as to represent how everyone out there sets them up in stereo?
So it's better to set it up like no one does.

 

[Old_School_Brad]

Think of it like this: your brain has a limited amount of focus for processing sound. The more speakers or sound sources you have, the less attention your brain gives to each one.

In a surround setup, your brain divides its focus across multiple speakers. For example, if you have seven speakers, your attention is spread out over 100% divided by e.g. 7. In a stereo setup, that same focus is split between just two speakers, so each one gets a lot more of your brain's attention. And with a single speaker, all of your brain’s attention is focused on that one source, which makes any flaws stand out more.

This is a very simplified way of looking at it, but perhaps it helps explain it more clearly.

 

[Vuki]

Think of it like this: your brain has a limited amount of focus for processing sound. The more speakers or sound sources you have, the less attention your brain gives to each one.

In a surround setup, your brain divides its focus across multiple speakers. For example, if you have seven speakers, your attention is spread out over 100% divided by e.g. 7. In a stereo setup, that same focus is split between just two speakers, so each one gets a lot more of your brain's attention. And with a single speaker, all of your brain’s attention is focused on that one source, which makes any flaws stand out more.

This is a very simplified way of looking at it, but perhaps it helps explain it more clearly.

Everything should be made as simple as possible, but not simpler.

 

[amirm]

How do you set up any loudspeakers as to represent how everyone out there sets them up in stereo?

On axis unless manufacturer says otherwise.

 

[Floyd Toole]

Think of it like this: your brain has a limited amount of focus for processing sound. The more speakers or sound sources you have, the less attention your brain gives to each one.

In a surround setup, your brain divides its focus across multiple speakers. For example, if you have seven speakers, your attention is spread out over 100% divided by e.g. 7. In a stereo setup, that same focus is split between just two speakers, so each one gets a lot more of your brain's attention. And with a single speaker, all of your brain’s attention is focused on that one source, which makes any flaws stand out more.

This is a very simplified way of looking at it, but perhaps it helps explain it more clearly.

A nice attempt to add clarification, but there is more. The reduced ability to hear resonant colorations in loudspeakers when more than one channel is operating is true only when multiple channels are radiating simultaneously. It is the large-room spatial information in recordings (envelopment - the sense of being in a large space) that is the distraction, not just the fact that there are multiple sources of sound. Listening room reflections are highly correlated with the loudspeakers located in the room. Recorded "reflections" of real or synthesized spaces are abstract quantities having no relationship to the space the eyes see. But these perceptions are what make stereo and multichannel so pleasurable.

There is no "envelopment" in mono signals. But, mono signals exist in multichannel recordings whenever there is a hard-panned image. The most blatant example is the center channel in movies, which does most of the important work, delivering most dialog and much on-screen action sounds. Solo instruments also appear in single loudspeakers. These are monophonic listening opportunities and this is why, even though there is an overall degradation in one's ability to hear resonance colorations in loudspeakers in stereo and multichannel recordings, in the end, listeners still prefer the most neutral loudspeakers. Hence, the recommendation to identify neutral loudspeakers in mono listening - or by interpreting useful anechoic data - and then enjoy them in stereo and multichannel. The more channels there are, up to a point, the more persuasive is the sense of envelopment. The added channels can deliver the appropriate long delayed large room reflections from appropriate directions. This is why loudspeaker directivity matters less in multichannel systems than it does in stereo or mono.

Does this help?

 

[Floyd Toole]

Maybe so, but the first graph also shows that for eq B stereo is much more preffered than mono.

There are no error-bars shown, but it is stated in the paper that the ONLY statistically signifcant differences were in the monophonic ratings. You might speculate about trends in the other data, but they are not proven facts in this set of experiments. This is discussed in detail in the upcoming 4th edition of my book.

 

[Floyd Toole]

I don't know about that paper, but the graph shows that greatest (and probably statistically significant) ddifferences are when eq method B is used, and then most listeners preffered stereo setup.
I bought 1st and 3rd edition of your book. It seems I would have to buy 4th also. Some discount would be appropriate especially talking in account that both of those books didn't help me much.

Sorry that the earlier editions of my book were not useful , but thank you for persevering.

For the 4th edition I reconfigured the data as shown here. The quality descriptions of the loudspeakers are related to the number of audible resonances - best being the most neutral sounding.

It makes it very evident that adding (simultaneously active) channels makes us less sensitive to resonant coloration in loudspeakers. Although only the mono ratings showed statistically significant differences, there are interesting trends to be observed. First, the "best" most neutral loudspeaker was highly rated in all comparisons. As channels were added the range of perceived sound qualities reduced, and the small differences between the sounds appeared to be differentiated more clearly, albeit not with statistical significance. More tests would have been required to obtain statistically significant differences with higher channel counts, but clearly the important point had been made. Clearly, hearing the differences is easier, and more sensitive when listening in mono, which is why it is the preferred technique.

This was a useful experiment, especially in that it showed persuasively that equalization of anechoic data can attenuate the audibility of resonances, and that equalization based on in-room measurements misses audible defects.

 

[Vuki]

Sorry that the earlier editions of my book were not useful , but thank you for persevering.

For the 4th edition I reconfigured the data as shown here. The quality descriptions of the loudspeakers are related to the number of audible resonances - best being the most neutral sounding.
View attachment 438502

It makes it very evident that adding (simultaneously active) channels makes us less sensitive to resonant coloration in loudspeakers. Although only the mono ratings showed statistically significant differences, there are interesting trends to be observed. First, the "best" most neutral loudspeaker was highly rated in all comparisons. As channels were added the range of perceived sound qualities reduced, and the small differences between the sounds appeared to be differentiated more clearly, albeit not with statistical significance. More tests would have been required to obtain statistically significant differences with higher channel counts, but clearly the important point had been made. Clearly, hearing the differences is easier, and more sensitive when listening in mono, which is why it is the preferred technique.

This was a useful experiment, especially in that it showed persuasively that equalization of anechoic data can attenuate the audibility of resonances, and that equalization based on in-room measurements misses audible defects.

Ok. Finally the right graph. Because the one @amirm showed just revealed that some speakers sound better in stereo than in mono.

 

[Vuki]

Looking once more at the last graph - it also shows that it's easier to notice finer differences between relatively good loudspeakers as the number of channels increases. Or it doesn't show anything relevant at all - it's quite dependent to interpretation.

 

[MattHooper]

Most people never get to experience a valid, blind, equal loudness comparison test.

It was being cognizant of some of your early work that compelled me, starting in the 90s, to do some of my own blind testing, meagre as it was. Since then I have blind tested things like cables and electronics… level matched, randomize, switching, etc… which often enough have exposed bias effects. I’d bet you would agree that personal experience with blind tests is really the best way for the power of bias effects to really sink in. If only all audiophiles had the chance.

The added channels can deliver the appropriate long delayed large room reflections from appropriate directions. This is why loudspeaker directivity matters less in multichannel systems than it does in stereo or mono.

Along those lines, I hope you don’t mind a question about the use of omnidirectional loudspeakers in multichannel surround systems.

This question arises from my own experience having owned MBL Radialstrahler omnis at one point. I do Sound Design for film/tv and of course our mixing studios use regular controlled directivity loudspeakers for the surround systems. That same for my own Home Theatre (7.0 surround).

But I did some early experimenting using the MBL omnis as left and right speakers in my surround set up, and I found the experience quite positive in some ways - listening to both some of my own work, as well as various well-known movies in surround. (For reference I was using THESE MBL OMNIS…in MY 2 channel / home theater room here … which allows for lots of flexibility in terms of controlling reflectivity).

Among the positives was the increased realism of acoustic space. For instance, the sense of the insect and rustling leaves layered backgrounds (and rain!) in the Jurassic Park movies just taking over the space in the room around the projected image just felt so realistic.

The main detriment I found was a lack of cohesion with my existing centre channel (a large, well designed W(TM)W configuration).
The imaging didn’t “ float free” as easily from my centre channel as it did from the Omnis.

But it seemed like it sure would be intriguing to be able to hear a full 5 or 7 channel surround system comprised of those omnis.
I’ve only read one review of an MBL surround system which reported the sensation of being in a realistic acoustic space to be fairly startling.

As I understand it, for surround sound, you have a preference for controlled directivity loudspeakers. (Better imaging, more accurate localization, and consistent timbre etc)

I’m wondering if you are aware of surround sound tests using only Omni directional loudspeakers. Or whether you have experienced such a system yourself, and if so, what you thought about the experience.

 

[kharan]

Sorry that the earlier editions of my book were not useful , but thank you for persevering.

For the 4th edition I reconfigured the data as shown here. The quality descriptions of the loudspeakers are related to the number of audible resonances - best being the most neutral sounding.
View attachment 438502

It makes it very evident that adding (simultaneously active) channels makes us less sensitive to resonant coloration in loudspeakers. Although only the mono ratings showed statistically significant differences, there are interesting trends to be observed. First, the "best" most neutral loudspeaker was highly rated in all comparisons. As channels were added the range of perceived sound qualities reduced, and the small differences between the sounds appeared to be differentiated more clearly, albeit not with statistical significance. More tests would have been required to obtain statistically significant differences with higher channel counts, but clearly the important point had been made. Clearly, hearing the differences is easier, and more sensitive when listening in mono, which is why it is the preferred technique.

This was a useful experiment, especially in that it showed persuasively that equalization of anechoic data can attenuate the audibility of resonances, and that equalization based on in-room measurements misses audible defects.

All the tests you show present the results by showing preferences for groups of listeners, which makes sense when you are trying to find speaker characteristics that are pleasing to the largest audience.

But across the tests that you have conducted, are there individuals that are consistently preferring a speaker with a specific colouration?

In more general terms, how much can individuals deviate from the statistical group preference?

 

[Floyd Toole]

All the tests you show present the results by showing preferences for groups of listeners, which makes sense when you are trying to find speaker characteristics that are pleasing to the largest audience.

But across the tests that you have conducted, are there individuals that are consistently preferring a speaker with a specific colouration?

In more general terms, how much can individuals deviate from the statistical group preference?

Most Harman listening tests were done using single listeners, always in the same location, to control room effects. Demos are different. Hundreds of listeners have been used over the years and with the exception of listeners with hearing defects, the ratings have been impressively consistent - i.e. all normal hearing listeners appear to prefer neutral loudspeakers in multiple comparison tests. Sean Olive tested this deliberately using a large number of listeners from different backgrounds: Olive, S.E. (2003). “Difference in Performance and Preference of Trained versus Untrained Listeners in Loudspeaker Tests: A Case Study”, J. Audio Eng. Soc., 51, pp. 806-825. In the end the loudspeaker ratings were very similar, but there were large differences in the consistency (variance) in the ratings. The result was interesting. The FL statistic is a measure of variability combined with the range of rating differences used (are the listeners certain about hearing the differences?). This is why we rely on listeners selected for "Normal" hearing and trained to hear resonances. The results are more trustworthy and they happen quickly.

However, one can ask a different kind of question. Assuming that the loudspeaker is free from timbral coloration due to audible resonances. what spectral balance is preferred? Here things get complicated because recordings are not consistent - the circle of confusion - and listening rooms dominate bass quality and that is about 30% of sound quality ratings. So, call it what you will: personal preference, room effects, recording differences, there will be variations in overall spectral balance. However, when tested, it is inexperienced listeners who seem to deviate most from neutral spectra - and some of the deviations are large. But starting with neutral loudspeakers allows one always to get back to something like what we might have been intended to hear - assuming that the room resonance problems have been handled. That is a separate story.

 

[Floyd Toole]

It was being cognizant of some of your early work that compelled me, starting in the 90s, to do some of my own blind testing, meagre as it was. Since then I have blind tested things like cables and electronics… level matched, randomize, switching, etc… which often enough have exposed bias effects. I’d bet you would agree that personal experience with blind tests is really the best way for the power of bias effects to really sink in. If only all audiophiles had the chance.



Along those lines, I hope you don’t mind a question about the use of omnidirectional loudspeakers in multichannel surround systems.

This question arises from my own experience having owned MBL Radialstrahler omnis at one point. I do Sound Design for film/tv and of course our mixing studios use regular controlled directivity loudspeakers for the surround systems. That same for my own Home Theatre (7.0 surround).

But I did some early experimenting using the MBL omnis as left and right speakers in my surround set up, and I found the experience quite positive in some ways - listening to both some of my own work, as well as various well-known movies in surround. (For reference I was using THESE MBL OMNIS…in MY 2 channel / home theater room here … which allows for lots of flexibility in terms of controlling reflectivity).

Among the positives was the increased realism of acoustic space. For instance, the sense of the insect and rustling leaves layered backgrounds (and rain!) in the Jurassic Park movies just taking over the space in the room around the projected image just felt so realistic.

The main detriment I found was a lack of cohesion with my existing centre channel (a large, well designed W(TM)W configuration).
The imaging didn’t “ float free” as easily from my centre channel as it did from the Omnis.

But it seemed like it sure would be intriguing to be able to hear a full 5 or 7 channel surround system comprised of those omnis.
I’ve only read one review of an MBL surround system which reported the sensation of being in a realistic acoustic space to be fairly startling.

As I understand it, for surround sound, you have a preference for controlled directivity loudspeakers. (Better imaging, more accurate localization, and consistent timbre etc)

I’m wondering if you are aware of surround sound tests using only Omni directional loudspeakers. Or whether you have experienced such a system yourself, and if so, what you thought about the experience.

As I describe in the 3rd edition - Section 7.4.6 - I purchased a pair of nearly omnidirectional Mirage M1s for my large, somewhat live, classical "concert hall" - the largest I could afford at the time . They did well in small room double-blind tests at the NRCC, and in this situation they were to simulate an orchestra: wide dispersion, lots of reflections, etc. combined with neutral timbre and very uniform dispersion (anechoic and in room measurements are shown). They worked well and did not draw attention to themselves. (I also had a home theater).

In conventional stereo systems, multidirectional loudspeakers seem to benefit this kind of music, but also even pop recordings with hard-panned L and R images - monophonic sounds. It is often considered more pleasing if the instrument or group of instruments does not localize to a single point in space. Local early reflections generate acoustical interference - the "dreaded" comb filtering - but there are many such filters even in small rooms, so they are rarely audible problems. Concert halls are just enormously complicated comb filters. Fortunately, humans have binaural hearing and instead of hearing coloration from the measured combs, we hear spaciousness and it is desirable.

I never think of "imaging" from surround loudspeakers, as only rarely and usually only momentarily are they required to deliver a hard-panned sound. The most critical requirement in my opinion is neutral timbre, so that they do not draw attention to themselves. So, an absence of resonances is fundamental, and wide dispersion is an option, about which I have no personal opinion. All serious multichannel tests I am aware of have used arrays of conventional forward-firing loudspeakers. What would be heard from multidirectional surround loudspeakers would depend on the reflectivity of the room boundaries, which, if reflective enough to make use of the "off axis" sounds might result in an overwhelming cacophony of early reflections. If envelopment is the desired effect - and I suspect it is - then one does not need a complex sound field. See section 15.7.1 in the 3rd edition. In those tests conventional 5 channel direct radiators did a splendid job of replacing many more channels.