Evidence-based Speaker Designs

[Frank Dernie]

Directivity is by orders of magnitude more important than (internal) cabinet design.

How many orders of magnitude?
Do you mean you have measurements showing the relative contribution of the drivers, fairly accurate. and the cabinet, signal related spurious, for a very large number of speakers and the listening tests quantifying how many orders of magnitude less important the spurious sound is?
Given the shape of a typical speaker, how much of the radiation to the side and rear from the cabinet contributes to the directivity?

 

[Frank Dernie]

Harman is known for doing listener based tests to drive how they design. And if their data says people like X, they give them X. If I knew everybody loved salt, I would overload all my stuff with salt. After the public loves it and starts becoming salt obsessed, it's only for a good reason. It just seems to be me like they are giving the public what the data shows they are most likely to like rather than having designers approach the problem with some pre-conceived preference first. But now that in itself is a preference.

If there are two speakers with similar directivity, one of which has a sophisticated cabinet radiating very little sound and a second where a significant proportion of the sound to the sides and rear is being produced by the cabinet not the drivers would they sound the same then?

 

[Rollomoto]

If there are two speakers with similar directivity, one of which has a sophisticated cabinet radiating very little sound and a second where a significant proportion of the sound to the sides and rear is being produced by the cabinet not the drivers would they sound the same then?

probably yes, for untrained ears, especially in non nearfield applications.

 

[Rollomoto]

How many orders of magnitude?
Do you mean you have measurements showing the relative contribution of the drivers, fairly accurate. and the cabinet, signal related spurious, for a very large number of speakers and the listening tests quantifying how many orders of magnitude less important the spurious sound is?
Given the shape of a typical speaker, how much of the radiation to the side and rear from the cabinet contributes to the directivity?

I would say 10 ;-)

Internal cabinet design – that ist what we are talking of – has nothing to do with directivity
The external of course has, that is size and shape or the front baffle (a waveguide as part of the baffle)

 

[Rollomoto]

When it comes to loudspeaker preferences, the Toole/Olive research produced evidence that listeners preferred non-constant CD. Indeed, the Olive preference rating system penalises CD designs, instead favouring designs with a downward-sloping off-axis response.

Most of Harmans designs that are based of Toole's work are CD (many Revel, JBL 30x, 70x, M2 ...). There is a preference for CD in listening tests, that is the reason why most professionals aim for it. It think @Floyd Toole states that very clearly.

But you are right in saying that CD and evidence-based design do not need to be the same thing. For CD you have to master directivity, it is much easier so build non-CD, that is what the industrie did in the last decades. Its especially hard if you extend CD to mids and bass (Dutch&Dutch, ME Geithain, Kii-Audio, Grimm...). So its quite a a quality criterion.

 

[Frank Dernie]

probably yes, for untrained ears, especially in non nearfield applications.

So the pioneering work done by Dudley Harwood et al on loudspeaker cabinets at the BBC research centre were actually just changing directivity and they didn't realise it.
Thank goodness that is cleared up at last.

 

[andreasmaaan]

There is a preference for CD in listening tests

Would you mind pointing me in the direction of those tests?

The main body of research I'm aware of is Olive's, which did not favour CD (arguably partly because it may not have been represented in the sample, but it seems we'll never know).

 

[napilopez]

@bobbooo
As I said the whitepaper you are fan of does not show any useful directivity measurements, since it only shows the tweeter up to 60° in horizontal dimension. Proper directivity plots show vertical und horizontal planes up to to 360°, like Neuman, Genelec and other professionals do.

Another example:

(taken from: https://ggntkt.de/en/model-m1/technische-daten/ )

The stereophile plots you are sharing again, do show that this Q-Acoustics is not constant directivity (CD), since it is to narrow into high frequency. To achieve CD you have to use a waveguide, especially in a 2-way. This is why Neuman, Genelec, JBL and other professionals use a complex shaped front baffles, based on the findings of Toole/Olive/Harman.

My impression:
Q-Acoustics is not true evidence-based design. It seems that marketing dictates the form factor and aesthetics and Mr. Fink had to make the best of it. That's a different approach from aiming for the best performance from the very beginning.

As for the digital/active discussion:
I see no contradiction to my statement – if you strive for max performance you have to go digital and active.

I agree with a @bobbooo and personally find the complaints levied against Q acoustics kind of odd. I like their speakers, so that's my bias.

We consider the majority of the Harman stereo portfolio to "evidence-based" yet they only publish measurements for a small minority of their speakers. If they do, it's usually pretty hard to find them. I don't think you can really sin Q acoustics for neglecting to provide that information given their speakers do, in fact, seem to measure pretty well within their price brackets.

I don't like the lack of detailed measurements either, but it would be far from the first time I've seen a hifi speaker with good measurements that are hard to find or non-existent.

With the exception of the flagship Concept 300, all of the company's bookshelf speakers cost less than $530 a pair. Somewhat ironically given your argument about directivity, I'd argue the strength of the $315 3020i and $500 concept 20 is actually how relatively well controlled their horizontal directivity is within their price brackets.

If you look at their measurements, the on-axis has a midrange scoop, but it largely remains consistent off axis, pointing to pretty good direcitvity control, and suggesting the dip can be at least partly EQ'd out if you want.

3020i:


Concept 20:


I think those are pretty good measurements for $315 and $530 a pair speakers, respectively. I only have vertical measurements out to 30 degrees for them, but within that window they seem about as well controlled as most of the other 2 ways I've tested.

The Concept 300 costs $4,500, which is pricey, but not at all outlandish by hi-fi standards, especially given it includes the fancy stands. It performs quite well. It's not the cheapest speaker with such good measurements, but it's far from the most expensive. The KEF Reference 1 costs $7500 for comparable horizontal measurements.

And as @andreaasman noted, constant directivity is just one way of designing speakers, and arguably not the best one. None of revels speakers are truly constant directivity as far as I know - that shows up as a level DI curve rather than a gently rising one.

In general, Dr. Toole seems to believe wider directivity is preferred to narrow directivity for recreational use at home (as opposed to for mixing engineers and some audio professionals who often preferr narrow directivity). In one part of his book he even suggests somewhat uneven wide directivity might be preferred to better-controlled narrow directivity (within reason).

Indeed, this is supported by the blind test that happened between the Revel Salon2 (wide directivity) and JBL M2 (constant directivity); the Revel won, despite having what I would consider slightly worse directivity (a dip around the crossover).

As for "striving for maximum performance" I also see @bobbooo's point too: every speaker has design constraints. If we were always "striving for maximum performance," every speaker would have to be large enough to hit crazy high SPLs, cost a bazillion dollars, and likely sacrifice aesthetics along the way. I also prefer active designs, but sometimes passive is simple more practical or cost effective. This does not stop them from being evidence based. And again, Harman's blind test champion is a passive speaker is a passive one, not an active one.

Also, while cabinets are a big part of Q-Acoustics' design, I don't see them marketing their drivers as being anything fancy. They're seemingly unremarkable paper cones.

As far as I'm concerned, once you have decent frequency response and directivity and the price is reasonable, everything else is fair game. So if Q Acoustics wants to make its "hook" controlling resonances, so be it.

Even though I do tend to believe that resonances shouldn't be very audible if they don't show up in the frequency response, based on what @Frank Dernie has said it seems Q Acoustics and others believe otherwise.

As a counterpoint to myself, while I wish Q Acoustics would correlate their work reducing resonances with blind tests, we know that the detection threshold low-q/wide-band resonances is incredibly low. So even if a resonance is not readily apparent in on and off axis frequency response graph, it doesn't mean it's not still audible. Per Toole, these are some detection thresholds for resonances.


I don't know about you, but a resonance like the one on the bottom right would be quite hard to spot in a frequency response graph.

Moreover, the above thresholds were determined in an anechoic environment, but he says resonances are likely to be more audible in a typical listening room because of the repetition.

So while we don't know just how much of an audible improvement reducing resonances in the hardware makes vs doing so in an active design, it's not hard to see why it can be useful to do so in the hardware like Q Acoustics does.

 

[napilopez]

Most of Harmans designs that are based of Toole's work are CD (many Revel, JBL 30x, 70x, M2 ...). There is a preference for CD in listening tests, that is the reason why most professionals aim for it. It think @Floyd Toole states that very clearly.

But you are right in saying that CD and evidence-based design do not need to be the same thing. For CD you have to master directivity, it is much easier so build non-CD, that is what the industrie did in the last decades. Its especially hard if you extend CD to mids and bass (Dutch&Dutch, ME Geithain, Kii-Audio, Grimm...). So its quite a a quality criterion.

You appear to be conflating constant directivity with controlled directivity. To be fair there aren't any "official" definitions as far as I know.

Still, in general, I consider constant directivity to be a more specific term meaning directivity that changes little over a wide range of frequencies. This is a speaker like the JBL M2 or Dutch and Dutch 8C. In a spinorama curve, it shows up as a very wide section of the DI curve with a roughly 0 slope, hence the "constant" part.

This is normally seen in speakers with large waveguides, although technically an omnidirectional speaker would also have constant directivity. It just means the directivity is similar over a wide range of frequencies. But in general for traditional box speakers, constant directivity designs are also relatively narrow directivity designs.

Revel speakers, on the other hand, fall under the controlled directivity camp. It just means care is taken to ensure a smooth change in directivity as you go up in frequency. These speakers normally have a gently rising DI curve. They also normally have small, shallow, or none-existant waveguides. Salon 2:

At no point in the frequency response is the directivity really "constant," it's just decently well controlled.

To your earlier point about the Q Acoustics being too narrow in the high frequencies, this does not follow the data. Revel speakers perform very similarly, and it is typical for wider-directivity designs to have a steep fall-off in the highest frequencies where the treble unit becomes more directional.

The Concept series, in fact, have very similar horizontal directivity characteristics to the Revel family.

Concept 300 at 45/60/75 degrees off-axis:

Here's Revel's M126be, $4,000 a pair, so when you factor in the stands it costs basically the same as the Concept 300:

If anything, the M126Be maintains slightly less energy into the top octave. The overall frequency response is also lower in level than the 0-degree measurement, which coupled with more widely-spaced lines points to slightly wider directivity for the concept 300.

Here's the Revel F206:

Also clearly a controlled, but not "constant" directivity design. So the Q Acoustics, like the Revels, fall under the controlled directivity/smoothly changing directivity camp.

EDIT: Lastly, an old comment from Dr. Toole himself to clear things up:

Cone/dome loudspeakers tend to show a gently rising directivity index (DI) with frequency, and well designed horn loudspeakers (like the M2) exhibit quite constant DI over their operating frequency range. There is no evidence that either is advantageous - both are highly rated by listeners.

That said, as mentioned before, Dr Toole has also said people tend to prefer wide directivity in recreational situations.

 

[andreasmaaan]

A large part of this thread seems to be devoted to constant directivity designs, whereas, to my knowledge, most direct evidence we have regarding listener preference does not support constant directivity as a design goal.

However, that's not to say the evidence is conclusive that CD designs are not preferred by listeners; it just seem to me we're at best jumping the gun here, and at worst actually mistaken.

I wonder if perhaps the technical elegance of CD partly explains our willingness to accept it?

Anyway, all I really wanted to do was getting a discussion going about evidence for or against constant directivity as a design goal...

FWIW, although the Olive work essentially finds against constant directivity, IMO there is some strong indirect evidence in support of it, mostly in studies looking at the precedence effect, which has been observed to be most effective when the spectra of direct and reflected sounds are similar (see for example Toole). Constant directivity speakers of course ensure that first reflections (which are not damped or absorbed) have the most similar spectrum to the direct sound. This would favour wide constant directivity for perceived spaciousness (but arguably not narrow constant directivity).

That's the best I've got, anyone else have any other evidence?

 

[goldark]

In above plot you can see their typical directivity problem around 3kHz due to their too small waveguide which they try to compensate with an on-axis dip.

I would argue that imperfect, but generally good measurements should not preclude a speaker design from being "evidence based." Doing so would lead to the absurd conclusion that only speakers with perfect measurements can be considered "evidence based" - not saying that's the conclusion you make, but I felt it needed to be said.

 

[Ilkless]

A large part of this thread seems to be devoted to constant directivity designs, whereas, to my knowledge, most direct evidence we have regarding listener preference does not support constant directivity as a design goal.

However, that's not to say the evidence is conclusive that CD designs are not preferred by listeners; it just seem to me we're at best jumping the gun here, and at worst actually mistaken.

I wonder if perhaps the technical elegance of CD partly explains our willingness to accept it?

Anyway, all I really wanted to do was getting a discussion going about evidence for or against constant directivity as a design goal...

FWIW, although the Olive work essentially finds against constant directivity, IMO there is some strong indirect evidence in support of it, mostly in studies looking at the precedence effect, which has been observed to be most effective when the spectra of direct and reflected sounds are similar (see for example Toole). Constant directivity speakers of course ensure that first reflections (which are not damped or absorbed) have the most similar spectrum to the direct sound. This would favour wide constant directivity for perceived spaciousness (but arguably not narrow constant directivity).

That's the best I've got, anyone else have any other evidence?

I had said indirect evidence in mind when linking to constant directivity, thanks for articulating it for me! And a point discussed earlier by another user is correct imo too, constant directivity is a more difficult target than merely controlled to engineer to; demands more technical competence and awareness of the manufacturer and that should be recognised. One might stumble on controlled directivity unintentionally, but I find it difficult to believe a constant directivity design is not deliberately so, and it reflects that the manufacturer is aware of the research literature at least. FWIW I would take a limb and say part of why CD is disfavoured is because of the compromises historically taken to get there - eg. tons of resonances

 

[andreasmaaan]

I had said indirect evidence in mind when linking to constant directivity, thanks for articulating it for me! And a point discussed earlier by another user is correct imo too, constant directivity is a more difficult target than merely controlled to engineer to; demands more technical competence and awareness of the manufacturer and that should be recognised. One might stumble on controlled directivity intentionally, but I find it difficult to believe a constant directivity design is not deliberately so, and it reflects that the manufacturer is aware of the research literature at least. FWIW I would take a limb and say part of why CD is disfavoured is because of the compromises historically taken to get there - eg. tons of resonances

Interesing point, yeh. I also simply wonder whether CD designs even figured at all in the Toole/Olive research?

I think we're definitely overdue for further preference studies of the scale of those that Toole and Olive, but with some more modern coaxial, cardioid, and possibly even vertical array designs in the mix. I think one of the major holes in the existing research is that, as far as it's possible to tell, virtually all the models tested were either conventional box speakers or panel dipoles.

In the meantime though, I'm reluctant to accept outright the idea that CD is a design goal that you could call "evidence-based", although like I said earlier, there's certainly an elegance to the idea that a speaker should have a flat frequency response at every possible point in space.

 

[napilopez]

A large part of this thread seems to be devoted to constant directivity designs, whereas, to my knowledge, most direct evidence we have regarding listener preference does not support constant directivity as a design goal.

However, I that's not to say the evidence is conclusive that CD designs are not preferred by listeners; it does seem to me we're at best jumping the gun here, and at worst actually mistaken.

I wonder if perhaps the technical elegance of CD partly explains our willingness to accept it?

Anyway, all I really wanted to do was getting a discussion going about evidence for or against constant directivity as a design goal...

FWIW, although the Olive work essentially finds against constant directivity, IMO there is some strong indirect evidence in support of it, mostly in studies looking at the precedence effect, which has been observed to be most effective when the spectra of direct and reflected sounds are similar (see for example Toole). Constant directivity speakers of course ensure that first reflections (which are not damped or absorbed) have the most similar spectrum to the direct sound. This would favour wide constant directivity for perceived spaciousness (but arguably not narrow constant directivity).

That's the best I've got, anyone else have any other evidence?

You actually bring up an interesting point with the precedence effect and what is "most similar" to the direct sound. I agree that "wide constant directivity" might be champion... to a certain degree of course. The ultimate wide constant directivity is omnidirectional, and I'm not quite ready to believe omnis are the best for all spaces and recordings.

That said, you could also argue that the wider a speaker's directivity, the more it approaches "wide constant directivity" (it would look like a smaller slope in the DI curve). It stands to reason the wider directivity the speaker, the more similar the reflected sound is to the on-axis sound, until you reach omni-directional (which again, is maybe sometimes too wide?).

So I'm willing to accept a bit of a "less pretty" DI curve on wide directivity speakers because a louder reflected sound seems to be associated with precedence and preference.

Edit: the nice bonus is that it such speakers also tend to be more room independent in my experience. On the other hand, the narrower the directivity the more you can "teak" the soudn by angling the speaker in different ways. And of course, I'm painting with a broad brush here and just speaker about trends. Some speakers are narrower at some frequencies and wider at others, or the directivity change might be different with respect to angles, etc.

 

[andreasmaaan]

You actually bring up an interesting point with the precedence effect and what is "most similar" to the direct sound. I agree that "wide constant directivity" might be champion... to a certain degree of course. The ultimate wide constant directivity is omnidirectional, and I'm not quite ready to believe omnis are the best for all spaces and recordings.

That said, you could also argue that the wider a speaker's directivity, the more it approaches "wide constant directivity" (it would look like a smaller slope in the DI curve). It stands to reason the wider directivity the speaker, the more similar the reflected sound is to the on-axis sound, until you reach omni-directional (which again, is maybe sometimes too wide?).

So I'm willing to accept a bit of a "less pretty" DI curve on wide directivity speakers because a louder reflected sound seems to be reflected with precedence and preference.

The thing is (at least in my interpretation of the research ), the reason that wide directivity designs tend to be preferred is that they maximise lateral reflections, which increase perceived image size and spaciousness (without significantly harming imaging, within reason).

However, for this perecived spatial widening to occur, there needs to be a low degree of interaural cross-correlation. In other words, the spectrum of the sounds arriving at each ear need to be as different as possible. This occurs most when (otherwise identical) sounds arrive laterally, since the listener's HRTF ensures that there is maximum interaural time difference (ITD) and maximum interaural level difference (ILD) between the sounds arriving at each ear. It occurs least when sounds arrive from in front, behind, above and below.

This is why I'm also reluctant to view omni as a design goal: a higher degree of the first-reflected sound arrives from front, up and down, and is therefore more highly cross-correlated, and will not tend to contribute to this sense of spaciousness. OTOH, I'm not aware of any research suggesting that these reflections, if timbrally identical (or similar) to the direct sound, have a negative effect on perceived timbre.

But to cut a long story short, I'd be very interested in any speaker with wide (up to 180°) horizontal constant directivity, narrower vertical directivity, and as little as possible rearward radiation. (I'm yet to see a design that gets very close to fulfilling these criteria FWIW).

 

[napilopez]

But to cut a long story short, I'd be very interested in any speaker with wide (up to 180°) horizontal constant directivity, narrower vertical directivity, and as little as possible rearward radiation. (I'm yet to see a design that gets very close to fulfilling these criteria FWIW).

This would be a very interesting design indeed. What would such a speaker look like? Perhaps an active speaker like the Kii three but with added emphasis on reducing extraneous vertical radiation?

 

[bobbooo]

However, for this perecived spatial widening to occur, there needs to be a low degree of interaural cross-correlation. In other words, the spectrum of the sounds arriving at each ear need to be as different as possible. This occurs most when (otherwise identical) sounds arrive laterally, since the listener's HRTF ensures that there is maximum interaural time difference (ITD) and maximum interaural level difference (ILD) between the sounds arriving at each ear. It occurs least when sounds arrive from in front, behind, above and below.

Have there ever been any speakers designs with intentionally asymmetrical lateral directivity in order to minimise this interaural cross-correlation? I'm thinking wide directivity on the 'outsides' of a stereo pair to maximise lateral reflections, but narrow directivity on the 'insides'. Of course a downside is this would also narrow the listener 'sweet spot'.

 

[andreasmaaan]

Have there ever been any speakers designs with intentionally asymmetrical lateral directivity in order to minimise this interaural cross-correlation? I'm thinking wide directivity on the 'outsides' of a stereo pair to maximise lateral reflections, but narrow directivity on the 'insides'. Of course a downside is this would also narrow the listener 'sweet spot'.

The Gedlee designs are designed for flat response 20° (or was it 30°) off-axis, and to be set up such that the forward wavefront hits the opposite side wall rather than the adjacent one, the idea being that the sweet spot is actually widened (SPL from the L channel increases as you rotate towards the R channel and vice versa).

In fact, I don't interpret the research as suggesting that reflections from the opposite side wall are less useful for perceived spaciousness than relfections from the adjacent ones, so I'm not sure that designing (or positioning) a speaker to minimise them would be of use.

I hope I understood your question correctly though?

 

[napilopez]

Have there ever been any speakers designs with intentionally asymmetrical lateral directivity in order to minimise this interaural cross-correlation? I'm thinking wide directivity on the 'outsides' of a stereo pair to maximise lateral reflections, but narrow directivity on the 'insides'. Of course a downside is this would also narrow the listener 'sweet spot'.

This is basically what the Polk SDA system does, no? The two midwoofers and tweeters on the "inside" cancel each other out, while the ones on the outside deliver a wider image. On the L800, the baffle is angles to further aid in this effect.

The bonus benefit is that aside from potentially helping with wide directivity, it also means that you don't get the inherent IAC stereo dip around 2Khz-ish, or at least less of it.

Of course, I don't know how well the system actually works. I have the L200 in for review (I'm very jealous of Amir's review pace) and I quite like it, but of course it has no SDA.

 

[andreasmaaan]

This is basically what the Polk SDA system does, no? The two midwoofers and tweeters on the "inside" cancel each other out, while the ones on the outside deliver a wider image. On the L800, the baffle is angles to further aid in this effect.

The bonus benefit is that aside from potentially helping with wide directivity, it also means that you don't get the inherent IAC stereo dip around 2Khz-ish, or at least less of it.

Of course, I don't know how well the system actually works. I have the L200 in for review (I'm very jealous of Amir's review pace) and I quite like it, but of course it has no SDA.

Do you know if there's any more detailed technical information on this available? The Polk website wasn't much help, so I'm still not exactly sure what this does or how it works.