omnidirectional loudspeakers = best design available

[BenB]

It's either one, zero, or infinite. Pretty sure it is one of those three.
I will run some simulations and get back to all y'all on this.

This may be a joke, but I see a lot of truth to it. If we are to approximate the dispersion of most sound sources as omni (even though there's evidence this is wrong), and if we want to re-create the experience of having those sound sources in front of us authentically, we need to have an omni speaker located where each source was located (relative to the listener). Having two separated omni speakers create a phantom center is not going to generate the same room interactions that a source actually in the center would.

I think I've read and considered the implications of each post in this thread (but not all links, etc). I'm fairly well convinced that a center channel speaker should have dispersion that's equal in the horizontal plane from +90 to - 90 degrees, and have no dispersion behind it, since any reflections from behind it should already be in the mix (either recorded authentically, or placed there by the sound designer). This thread has had little attention paid to vertical dispersion, and whether it should be broad or limited or if it really matters, and if so how. I have thoughts on that from prior experience and research. I'm not sure what dispersion stereo speakers should have ideally, even though that's the primary focus. Based on this thread, I'm also not sure whether when I listen to that center channel, I should have my walls treated or not. I suppose a speaker designer need not concern himself with such questions as wall treatments.

I will suggest the following thought experiment (to all thread participants):

Consider going to your favorite venue, and recording your favorite live performance with a single omni-directional microphone, placed in your preferred location. All the reflections of the venue will be captured with proper levels and delays.
Now consider playing back that recording on a single speaker (your dream speaker), and listening to it in an anechoic chamber.
How do you think you would like the sound?

 

[Benedium]

Not sure if relevant but... Would an orchestra of speakers, each representing a single instrument, sound the same to a single listener as a pair of speakers? Would be a nice art project if not one for science.

 

[ahofer]

I just want to say I don’t believe you can spatially differentiate the keys of a piano in playback. Not even the *strings*, unless recorded with stereo mics close over the soundboard or across the keyboard.

 

[whazzup]

Do you have a second pair of speakers which you could easily move into the same room and same system as your main speakers?

What I have in mind is, place the secondary speakers on the floor behind the main speakers, facing up, hopefully with the main speakers shielding the secondary speakers so that they do not have a direct sound path to the listening area. Then listen to both sets simultaneously, vs just your original speakers, ideally adjusting the volume level such that the secondary speakers don't convey an unfair SPL advantage.

This would be far from an ideal polydirectional setup, and I have no idea what your system is like so I have no idea whether it's even feasible.

If this does what I expect, the loudspeakers will be less apparent as the sound source with the secondary speakers on, and you'll have a bit more sense of envelopment in the acoustic space on the recording. Timbre may or may not be improved. The tonal balance may become overly thick in the upper bass and lower midrange region. I also expect image specificity and overall clarity to be degraded... optimization isn't practical for this experiment.

Let me know if you have any questions.

And, thank you for considering this, even if you decide against it.

Just saying that I concur with your observations when I did my ghetto desktop setup with used speakers to try to achieve larger sound.
https://www.audiosciencereview.com/...all-rooms-fire-your-speakers-backwards.14968/

Direct front firing sound is sharp but 'narrow', sounds like coming out from my display panel plane, but with the rear firing speakers the sound becomes 'thicker', extending from behind the screen. I can even turn off the front speakers and enjoy the indirect sound. But of course, the imaging is not as pinpoint sharp. And all these are in context to my room and placement.

 

[pozz]

Not sure if relevant but... Would an orchestra of speakers, each representing a single instrument, sound the same to a single listener as a pair of speakers? Would be a nice art project if not one for science.

What you're nearing towards is a concept called spatial sampling. The best you'll get from audio there is wavefield synthesis.

There have been a lot of attempts to model orchestras using just speakers. Here's a recent example: https://www.researchgate.net/publication/236169414_A_loudspeaker_orchestra_for_concert_hall_studies

 

[Duke]

Your words, “the approach preferred around here”, is actually the approach that listeners prefer, coming out of the research.

Has "the research" you refer to investigated in depth any loudspeaker topologies wherein the off-axis energy spectrally matched the on-axis energy? If so, I'm unaware of it. My impression is that the research concentrated on direct-radiator cone-n-dome speakers, whose off-axis energy inevitably is rolled off (due to driver beaming) relative to its on-axis energy. The research may have uncovered the best way to design such a speaker, but if it hasn't adequately evaluated other topologies, then maybe it's not the last word on every aspect of the subject?

Quoting from Sean Olive's paper:

"LIMITATIONS OF MODEL

"The conclusions of this study may only be safely generalized to the conditions in which the tests were performed. Some of the possible limitations are listed below.

"1. Up to this point, the model has been tested in one listening room.

"2. The model doesn't include variables that account for nonlinear distortion (and to a lesser extent, perceived spatial attributes).

"3. The model is limited to the specific types of loudspeakers in our sample of 70." [emphasis Duke's]

I will respond to the rest of your post later. My recollection of what Toole wrote about the M1's in his large room is different from what you describe, so I need to do my homework before I respond.

 

[Newman]

You still misrepresent it as “the approach preferred around here”, as if this is some cultist corner of the discussion. My point.

Perhaps you should have referred to it as “the approach preferred by all the best available research evidence”?

Or show something better.

I agree that all research is non-universal, but until you have something better, it’s better not to make a bunch of counterclaims.

Also check my other post about directivity of voice and instruments. What do you think the room response is like, for such original source directivity?

P.S. Your post that I quoted also described “the approach preferred around here” as having a room response gently downward-sloping. As if the speaker directivity was causing that? Not true. Room responses *are* gently downward sloping. It’s in the nature of real-life rooms with any furnishings at all, or any performance venues with any upholstered seats, and more so with bodies in those seats. You say “I find merit in minimizing the spectral discrepancy between the direct and reflected sound”, but that is saying you find merit in echo chambers. Because that is the only way to minimise it.

cheers

 

[Duke]

You still misrepresent it as “the approach preferred around here”, as if this is some cultist corner of the discussion.

Poor choice of words on my part.

Perhaps you should have referred to it as “the approach preferred by all the best available research evidence”?

How about, "the approach preferred in Harman's listening tests."

P.S. Your post that I quoted also described “the approach preferred around here” as having a room response gently downward-sloping. As if the speaker directivity was causing that?

Okay, how about "estimated in-room response". That wording is taken from one of the graphs in Amir's loudspeaker measurement posts.

The "estimated in-room response" is a summation which is dominated by the off-axis response, and its gentle downward slope is indeed caused by speaker directivity. The room's acoustics are a separate topic, my bad for using the wording "room response" instead of "estimated in-room response."

You say “I find merit in minimizing the spectral discrepancy between the direct and reflected sound”, but that is saying you find merit in echo chambers. Because that is the only way to minimise it.

If you really do not see how having the off-axis response more closely match the on-axis response minimizes the discrepancy between the direct and reflected sound, let me know and I will try to explain it. If you're just trying to make me look stupid by putting absurdities in my mouth ("that is saying you find merit in echo chambers")...

 

[Duke]

And, as for Toole’s Mirage M1 that you (repeatedly in this thread) mention, you are leaving out critical context: he wrote that he had (mis?)-conceived of a listening room in his home that turned out to be horrendous, virtually a giant echo chamber that just plain sounded bad, and his domestic situation made acoustic treatment impossible. And when he tried to use the room to play classical music of the time which, he said, was too-often horrendously poorly recorded and placed large sections of the orchestra wholly in one speaker, he needed a loudspeaker that was really bad at soundstage reproduction and greatly blurred the sonic image. Enter the Mirage M1. Which he promptly left behind when he sold the house, never intending to have such a room again. Context.

I disagree with most of what you wrote here. My source is the 3rd edition of Toole's book, pages 189 - 193. I'm going to quote you and then quote Toole:

What Newman wrote: "[Toole] wrote that he had (mis?)-conceived of a listening room in his home that turned out to be horrendous, virtually a giant echo chamber that just plain sounded bad, and his domestic situation made acoustic treatment impossible."

What Toole actually said in his book: "I concluded that two stereo loudspeakers needed the largest concert hall I could afford at the time... Acoustically, it was a pleasant space to be in."

* * *

What Newman wrote: "... [Toole] needed a loudspeaker that was really bad at soundstage reproduction and greatly blurred the sonic image. Enter the Mirage M1."

What Toole actually said in his book: "The human precedence-effect processor must have been especially pleased [with the Mirage M-1's in that room] because the soundstage and imaging were surprisingly intact, and the room became a seamless extension of it, embellishing the envelopment illusion that served the intended function - classical music - very well. Such a venue was a luxury, but it turned out to be not only for the classical repertoire."

* * *

What Newman wrote: "Which he promptly left behind when he sold the house, never intending to have such a room again."

What Toole actually said in his book: "I was motivated to explore the extent to which such an experience can be imitated in the small-room multi-channel system in Figure 7.18 [a much smaller room in the same house]." In other words, he liked what was happening with the bipolar Mirage speakers in his big room so much that he subsequently worked to imitate the experience in his small multi-channel room.

What actually happened when Toole sold that house: "The exploration continued in my new listening room in California, a very different room." He then goes on to describe his multichannel music system and room, which I can describe if you'd like, but my point is that the Mirages started him on the path to where his personal system is today, which includes a room with minimal early reflections AND enhanced late reflections via "tasteful upmixing [to surround channels] for many recordings to embellish the sense of space."

* * *

Newman, some of the discrepancies between what you wrote and what Toole said in his book seem pretty large to me. I don't know what your source is, but would like to. Was it Toole, or someone else's take on Toole?

If you question the validity of my Toole quotes, let me know.

 

[Newman]

I have the same book, and I used it to write my original comment, and I stand by it. Bad room + bad classical recordings => blurring and expansion of the soundstage needed => M1 speakers. “A parade of loudspeakers went through that room, and all disappointed.” i.e. speakers that didn’t blur the image and had precision and worked well in good audio spaces, were no good here. In essence, it was a special case, for which he found an acceptable solution. He never repeated the room, and never returned to the speakers, or similar. Totally walked away from both. Even left the speakers in the house, i.e. gave them away. That’s what I call learning from experience. Not sure what you are taking from it? Possibly the wrong idea, that near-omni performance is ideal? The science has moved on from that, courtesy of the same writer, and numerous others. Off-axis side wall reflections *do* need to match direct sound quite closely, but not at the expense of creating directivities that are wildly different from actual voice and instrument directivities. Doing that is to divert away from naturalness, even towards blurring and artificial expansion of the soundstage.

What he was trying to say, in his passage about his experience with that room, was that you can kind of make all sorts of non-ideal rooms work, by carefully selecting compensating compromises.

When he wrote “acoustically it was a pleasant space to be in”, that does not refer to music playback. He was talking about social and personal ‘space’ of it.

 

[Newman]

The "estimated in-room response" is a summation which is dominated by the off-axis response, and its gentle downward slope is indeed caused by speaker directivity. The room's acoustics are a separate topic, my bad for using the wording "room response" instead of "estimated in-room response."

You miss my point. I don’t care about the word ‘estimated’ being added to your text. The ‘gently-downward slope’ of the (estimated or measured, you choose) room response is in the nature of the room and would still be there if the speaker was omnidirectional.

If you really do not see how having the off-axis response more closely match the on-axis response minimizes the discrepancy between the direct and reflected sound, let me know and I will try to explain it. If you're just trying to make me look stupid by putting absurdities in my mouth ("that is saying you find merit in echo chambers")...

OK, what sort of room do you think will not have a gently-downward-sloping response? An echo chamber is the only one I can think of, hence no absurdity intended.

cheers

 

[Duke]

I have the same book, and I used it to write my original comment, and I stand by it.

Wow.

Let's let anyone who is interested enough read what Toole actually wrote about his big room, about the Mirage M-1 in that room, and about what he subsequently did in other rooms, and draw their own conclusions.

Newman, can you show me where Toole says that room was the sonic disaster you allege, and where the M-1's were likewise? Does he actually say it, or are you reading in between the lines?

If the following long quote is a copyright violation, please somebody let me know and I'll take it down:

Edit: Taken down.

 

[Duke]

You miss my point. I don’t care about the word ‘estimated’ being added to your text. The ‘gently-downward slope’ of the (estimated or measured, you choose) room response is in the nature of the room and would still be there if the speaker was omnidirectional.

OK, what sort of room do you think will not have a gently-downward-sloping response? An echo chamber is the only one I can think of, hence no absurdity intended.

And you miss my point. Or maybe I have simply failed to state it clearly.

I'm talking about the spectral balance of the loudspeaker's off-axis response, NOT what the room does to it. Unless I misunderstand, the Harman target curves call for a flat on-axis response and a gently downward-sloping summed off-axis response. The in-room response can be substantially predicted from the spin data.

And I find merit in minimizing the spectral discrepancy between the on-axis response (or more precisely the direct sound) and the summed off-axis response. Please don't restate this as some absurdity.

The greater the spectral discrepancy between the first-arrival sound and the subsequent reflections, the less effective the human precedence-effect processor is. Do you agree with that?

Not sure if relevant but... Would an orchestra of speakers, each representing a single instrument, sound the same to a single listener as a pair of speakers? Would be a nice art project if not one for science.

Would not sound the same!

Finnish acoustican Tappio Lokki (no relation to the step-brother and arch-enemy of the God of Thunder) uses a virtual orchestra of loudspeakers, one for each instrument, to study the acoustics and psychoacoustics of concert halls:

 

[Newman]

Let's let anyone who is interested enough read what Toole actually wrote about his big room, about the Mirage M-1 in that room, and about what he subsequently did in other rooms, and draw their own conclusions.

If the following long quote is a copyright violation, please somebody let me know and I'll take it down:

That is not fair use, suggest you take it down.

 

[Duke]

That is not fair use, suggest you take it down.

I read the copyright notice inside the book and I think you're right.

So now I guess it's your word against mine as to what Toole actually said, because I can't prove he DIDN'T say what you claim by quoting the relevant part of the section. Third edition of Sound Reproduction, the Acoustics and Psychoacoustics of Loudspeakers and Rooms, Section 7.4.6, "Observations of an Audio Enthusiast - Toole (2016)", pages 188 to 193.

Here are a couple more things you say, in post 210, that Toole neither says nor implies:

"Bad room + bad classical recordings => blurring and expansion of the soundstage needed => M1 speakers." Nope.

"speakers that didn’t blur the image and had precision and worked well in good audio spaces, were no good here. In essence, it was a special case, for which he found an acceptable solution." That room was a challenging special case for which he found a solution good enough that he sought to imitate aspects of it in two subsequent rooms. You seem to think it's a bad thing for a speaker to work well in a challenging room, and I think it's an indication that a speaker is doing something right. Toole neither said nor implied that the M1 was "merely acceptable"; quite the contrary. In fact he said it performed well in the small listening room at the NRC, which led to him trying it in his big room.

I'm still interested in your answer to the question I asked in post 213:

"The greater the spectral discrepancy between the first-arrival sound and the subsequent reflections, the less effective the human precedence-effect processor is. Do you agree with that?"

Also from your post 210:

"Off-axis side wall reflections *do* need to match direct sound quite closely, but not at the expense of creating directivities that are wildly different from actual voice and instrument directivities. Doing that is to divert away from naturalness, even towards blurring and artificial expansion of the soundstage."

I agree with this, but based on your wording I'm not sure you understand the role the precedence effect plays in the perception of directivities.

 

[oivavoi]

Might as well kill 3 birds with one stone.

If I stand in front of you and talk, my voice is not anything like omnidirectional, and if I play a violin or guitar or clarinet while standing there, its music is also not omnidirectional, not even close. So, if an omnidirectional speaker is placed where I am standing and plays a recording of my speech and instrumental playing, and redirects all the sound omnidirectionally, that loudspeaker is performing a gross reproductive inaccuracy. An act of high unrealism.

The whole idea that omnidirectional playback is correct, or ideal, is a myth.

Here are the actual directivity vs frequency plots of voice and various instruments, © Toole:

View attachment 104251

The fact that they are not all the same means that no one loudspeaker can be perfect at reproducing all their directionalities. However, in broad and general terms, the speaker type that does best at reproducing the directionality of some of the most critical original sources such as voice, string and wind instruments, is ..... your typical front-firing loudspeaker that trends to omnidirectional in the bass!

(P.S. An omnidirectional speaker is a flat line on the zero axis at the bottom of the above chart. Pretty bad eh?)

So, if we compare the human voice directivity curve, above, to some forward-firing loudspeakers © Toole:

View attachment 104252

Again, imagine an omnidirectional loudspeaker on the above graph as a flat line on the zero axis. We can easily see that various conventional loudspeakers on the graph are far superior to an omnidirectional at reproducing something approaching the directivity of a human voice in a natural way. Furthermore, if we imagine adding the various other instruments from the top graph onto the bottom chart, we see the conventional loudspeakers even better at matching their directivities, and the omnidirectional even worse again. In fact, the omnidirectional is completely uncompetitive at the one feature that has been promoted as its natural advantage.

Thus we see that all the above arguments that idealise the directivity of omnidirectional loudspeakers, are based on a false premise, which changes all your conclusions.

Cheers

Thanks, good arguments!

I for one am fully aware that most indvidual musical instruments don't radiate omnidirectionally. The most thorough work on this topic is probably this I think: http://lib.tkk.fi/Diss/2011/isbn9789526042916/isbn9789526042916.pdf

But when one combines many musical instruments in an ensemble or orchestra the end result will often be more towards omnidirectional. Not completely and fully omni, but to a certain degree. This is because many instruments radiate in different directions which complement each other (it's not as simple as front back, except for some brass instruments), and that the wings of the orchestra often face towards the centre/director/each other, not towards the public.

This is fully subjective, but I have many times in my life practiced with acoustic ensembles, and when going around them in circles the sound seems to me to be much more constant and omni than from conventional loudspeaker systems. Others may disagree. Robert Greene, top violinist and hifi guy, famously prefers narrow directivity speakers. But for me at least, omni speakers feel closer to how real instruments would interact in my home.

But the main argument for why omnis can make sense is actually not about how it may or may not mimic the directivity of instruments, but about how the indirect sound becomes more similar to the direct sound. That's the heart of it. The point is still to listen to the direct sound - as with conventional speakers - but that the indirect soundfield becomes more spectrally correct. This makes the loduspeakers disappear more easily, and there are indications that it may enhance our spatial perception (re "the second look" hypothesis Indescribed before).

 

[andreasmaaan]

I'm critical of claims that omnidirectional (or wide-directivity) speakers are inherently lower-fidelity than conventional monopoles because the "added" in-room reflections they produce create some kind of (possibly euphonic) "distortion".

For those making such an argument, there seem to me to be only two possible frameworks:

1. The idea that any reflection is a form of distortion, thus the lower in level reflections are, the higher the fidelity.
2. The idea that, although reflections are not inherently distorting, there is some "most correct" DI for a loudspeaker (or perhaps some maximum tolerable ratio of direct to reflected sound) such that any level of reflected sound that exceeds this ideal constitutes "distortion".

The problem with (1) is that it implies that the narrower the directivity of the speaker and the more absorptive the room, the higher the fidelity of the reproduction. Its logical conclusion must be that only an anechoic environment is suitable for the highest possible fidelity playback. That's an interesting idea, but I don't think it's what most people have in mind when they say that omni speakers are lower-fidelity than conventional speakers.

The problem with (2) is that any particular choice of loudspeaker DI or room acoustics must be made with reference to information that is not present in the signal. Recordings just do not contain information about the off-axis response of the loudspeaker or the acoustical properties of the listening room.

In fact, it could be argued that constant-directivity speakers (of which omnidirectional speakers are just a special case) better satisfy the requirements of fidelity to the recording, because at any particular off-axis location, the frequency response is flat. And by corollary, the spectrum of every reflection from a constant-directivity (including omnidirectional) speaker is also more fidelitous to the recording than for a conventional speaker (which does not have a constant DI).

Note I’m not arguing here that omni speakers are likely to be most preferred (which is a separate question). But I am arguing that a speaker that accurately reproduces the input at every point in space is surely higher-fidelity than a speaker that inaccurately reproduces the input at all but one point in space.

 

[Sancus]

Note I’m not arguing here that omni speakers are likely to be most preferred (which is a separate question). But I am arguing that a speaker that accurately reproduces the input at every point in space is surely higher-fidelity than a speaker that inaccurately reproduces the input at all but one point in space.

This actually seems like a pretty wild assumption to make. Who is creating any recording such that they expect it to be reproduced on a system that does this? Especially since such a system doesn't actually exist.

Recordings just do not contain information about the off-axis response of the loudspeaker or the acoustical properties of the listening room.

Not completely true; sometimes they do contain time/phase shifted ambience like reverb etc. Which is sometimes actually recorded from the room, or other times deliberately added during mixing. One could argue that something like an Auro3D recording, which can have more channels dedicated to reflections than direct sound, has quite a lot of information about the acoustical properties of the listening room, if indirectly.

In fact, it could be argued that constant-directivity speakers (of which omnidirectional speakers are just a special case) better satisfy the requirements of fidelity to the recording, because at any particular off-axis location, the frequency response is flat.

The sound of a narrow dispersion CD loudspeaker and a true omni are going to be very different, so how can you argue both satisfy some ill-defined fidelity requirement?

 

[andreasmaaan]

This actually seems like a pretty wild assumption to make. Who is creating any recording such that they expect it to be reproduced on a system that does this? Especially since such a system doesn't actually exist.

It’s not an assumption, it’s a fact. The recording contains information about amplitude, frequency, and phase; nothing more.

It’s only when you begin contemplating the intentions or expectations of the creators that you introduce assumptions.

Not completely true; sometimes they do contain time/phase shifted ambience like reverb etc.

Yes, but time/phase shifted ambience and reverb etc. do not contain information about the off-axis response of the loudspeaker on which they will be reproduced or the acoustical properties of the reproduction room (although of course they may indirectly contain such information about the objects recorded or the acoustical properties of the recording space).

 

[oivavoi]

I'm critical of claims that omnidirectional (or wide-directivity) speakers are inherently lower-fidelity than conventional monopoles because the "added" in-room reflections they produce create some kind of (possibly euphonic) "distortion".

For those making such an argument, there seem to me to be only two possible frameworks:

1. The idea that any reflection is a form of distortion, thus the lower in level reflections are, the higher the fidelity.
2. The idea that, although reflections are not inherently distorting, there is some "most correct" DI for a loudspeaker (or perhaps some maximum tolerable ratio of direct to reflected sound) such that any level of reflected sound that exceeds this ideal constitutes "distortion".

The problem with (1) is that it implies that the narrower the directivity of the speaker and the more absorptive the room, the higher the fidelity of the reproduction. Its logical conclusion must be that only an anechoic environment is suitable for the highest possible fidelity playback. That's an interesting idea, but I don't think it's what most people have in mind when they say that omni speakers are lower-fidelity than conventional speakers.

The problem with (2) is that any particular choice of loudspeaker DI or room acoustics must be made with reference to information that is not present in the signal. Recordings just do not contain information about the off-axis response of the loudspeaker or the acoustical properties of the listening room.

In fact, it could be argued that constant-directivity speakers (of which omnidirectional speakers are just a special case) better satisfy the requirements of fidelity to the recording, because at any particular off-axis location, the frequency response is flat. And by corollary, the spectrum of every reflection from a constant-directivity (including omnidirectional) speaker is also more fidelitous to the recording than for a conventional speaker (which does not have a constant DI).

Note I’m not arguing here that omni speakers are likely to be most preferred (which is a separate question). But I am arguing that a speaker that accurately reproduces the input at every point in space is surely higher-fidelity than a speaker that inaccurately reproduces the input at all but one point in space.

Great minds etc Fully agree. Constant directivity is the goal, whether that is accomplished with wide or narrow or omni or bi-directional speakers is a matter of preference, musical taste and room acoustics, as I see it. I like both omnis and narrow-dispersion speakers, particularly of the large horn variety. Ideally I would have liked to have both.