Evidence-based Speaker Designs

[thewas]

I guess the problem of constant directivity is that most existing such loudspeakers aren't really fully so, but start having CD behaviour usually from some mid region frequency. To approximate CD you need usually a cardioid woofer/mid like the D&D 8c:

 

[bobbooo]

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?

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.

I was thinking more something like the below image, but with the waveguide extending further to the left of the tweeter to cover the entirety of the black oval surround (and also possibly extending less to the right of the tweeter), resulting in a horizontally asymmetric waveguide with wider dispersion on the left than the right. If this were used as a left channel speaker, this would mean maximal lateral reflections off the left wall, but minimal interaural crosstalk radiated to the right ear. When paired with a right channel speaker with mirrored geometry, this would result in wide directivity on the 'outsides' of the stereo pair to maximise lateral reflections, but narrow directivity on the 'insides' to minimise crosstalk. (Interestingly, looking at the waveguide, it seems like this speaker was designed to have wide horizontal but narrow vertical directivity as you suggested @andreasmaaan.)

 

[thewas]

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.

Yes, maybe I should have clarified that better, on the other hand we are in 2020 and there are tons of loudspeakers in the market with just ok-ish horizontal directivity (not to talk about the vertical one) so if we would include that one on evidence based we would have to include a hell of a list.
Anyway such a categorisation like most is in the end subjective so unless we clearly define and all agree what we mean with "evidence based" there will be different views.

 

[andreasmaaan]

I was thinking more something like the below image, but with the waveguide extending further to the left to cover the entirety of its black surround, resulting in a horizontally asymmetric waveguide with wider dispersion on the left than the right. If this were used as a left channel speaker, this would mean maximal lateral reflections off the left wall, but minimal interaural crosstalk radiated to the right ear. When paired with a right channel speaker with mirrored geometry, this would result in wide directivity on the 'outsides' of the stereo pair to maximise lateral reflections, but narrow directivity on the 'insides' to minimise crosstalk. (Interestingly, looking at the waveguide, it seems like this speaker was designed to have wide horizontal but narrow vertical directivity as you suggested @andreasmaaan.)

The thing is, the difference in angle between left and right ear is more or less negligible in relation to the speaker. If you're seated at a typical listening distance of 2-3m, the outer ear is only going to be 5 or 6° further off-axis than the ear closer to the speaker. Even nearfield, say 1m, the difference is only going to be less than 15°. So I'm not sure you can really reduce interaural crosstalk to any meaningful extent using a waveguide in this way. Or at least, I can't imagine how it could. Unless I'm missing something?

The other issue I foresee is that the lowest frequency to which an asymmetrical waveguide like this will control directivity is going to be higher on the narrow-angled side than on the wide-angled side since, in a constant directivity waveguide, the lowest frequency of operation is inversely related to coverage angle.

 

[bobbooo]

The thing is, the difference in angle between left and right ear is more or less negligible in relation to the speaker. If you're seated at a typical listening distance of 2-3m, the outer ear is only going to be 5 or 6° further off-axis than the ear closer to the speaker. Even nearfield, say 1m, the difference is only going to be less than 15°. So I'm not sure you can really reduce interaural crosstalk to any meaningful extent using a waveguide in this way. Or at least, I can't imagine how it could. Unless I'm missing something?

It's not just the direct sound from the left speaker to the right ear that would be minimised with such an asymmetrical waveguide - there would be less reflected sound from the left speaker via the right wall reaching the right ear too (and of course the opposite for the right speaker / left ear). Whether all this would be perceptually preferable however is uncertain.

The other issue I foresee is that the lowest frequency to which an asymmetrical waveguide like this will control directivity is going to be higher on the narrow-angled side than on the wide-angled side since, in a constant directivity waveguide, the lowest frequency of operation is inversely related to coverage angle.

That's assuming we want constant directivity, which I thought we concluded there's no actual hard evidence is better

As Floyd Toole said in 2018:

Almost 50 years of double-blind listening tests have shown persuasively that listeners like loudspeakers with flat, smooth, anechoic on-axis and listening-window frequency responses. Those with smoothly changing or relatively constant directivity do best.

So smoothly changing (i.e. 'controlled') directivity is just as preferred as constant directivity. Also, note he says relatively constant, so brands chasing perfect CD may actually be barking up the wrong tree - it may turn out that even within the group of people who prefer this type of directivity, there's actually an optimum medium of directivity uniformity above which preference drops off.

 

[andreasmaaan]

That's assuming we want constant directivity, which I thought we concluded there's no actual hard evidence is better

Ah, poor choice of words on my part. The points I mentioned apply to any waveguide with constant directivity on any particular axis. So what I meant was, if the coverage angle of a waveguide is asymmetrical, the lowest frequency to which the waveguide controls directivity will be higher on the side with a narrower coverage angle.

But having thought this through further, I'm now starting to doubt my earlier assumptions. Basically, I was applying Keele's formula for CD waveguides:

F = K / (a*w)

where:
F = frequency at which the waveguide ceases to control directivity (Hz)
K = 25,306 (unitless, experimentally derived)
a = waveguide included angle in degrees
w = waveguide mouth width (or height) in metres

So if the coverage angle is narrower on one side (all else equal), the lowest frequency to which the waveguide controls directivity will be higher on that side.

To plug in an example, imagine a waveguide of width 10cm with a coverage angle of 60° on the right hand side and 30° on the left hand side.

For the 60° side, F = 25,3000 / (2*60*0.1) = 2,108Hz

For the 30° side, F = 25,3000 / (2*30*0.1) = 4,217Hz

In other words, all else equal, the waveguide will control directivity down to one octave lower on the side with double the included angle.

Or at least, that's how I imagined it would work at first glance. Now I'm less sure, in fact I think what would happen would be a bit more complex, with a difference in F on either side, but not as great as a 1:2 ratio.

As Floyd Toole said in 2018:
So smoothly changing (i.e. 'controlled') directivity is just as preferred as constant directivity. Also, note he says relatively constant, so brands chasing perfect CD may actually be barking up the wrong tree - it may turn out that even within the group of people who prefer this type of directivity, there's actually an optimum medium of directivity uniformity above which preference drops off.

Yeh, what Toole says makes intuitive sense. I just don't see it in the research findings themselves.

Perhaps Toole has access to unpublished data that contradicts Olive, or perhaps he's just speculating, I dunno. But the clear conclusion of the Harman studies was that a downward sloping in-room/early reflections/power response was preferred.

There's even a specific formula prescribed for the magnitude of the downward slope in Olive's paper, such that it's possible to calculate how far down the in-room/early reflections/power response of the "ideal" speaker should be from the axial response at any given frequency (for example, the early reflections should be 10dB down at 16kHz).

In other words it's all very specific, and it very specifically concludes in favour of non-CD speakers

 

[thewas]

Very true, would be really interesting to directly compare a classic increasing directivity design with the continuous and around -10dB drop at high frequencies at the listeners position to a CD design with less and more linear FR there, I have the feeling the first one will be prepared and was also one of the reasons that the Revel was preferred to the M2. Unfortunately I had only had very few loudspeakers with CD (tweeter) till now in my place, but never kept any very long as somehow I didn't prefer their sound, but it can be just a coincidence because of the too small sample.

 

[tuga]

This thread (and to a lesser extent this forum as a whole) seems heavily biased towards active speakers, yet if we take a look at the speakers measured so far on here, there are an equal number of passive and active models in the top six when used with a sub according to preference rating. And this preference rating ranks according to performance that is firmly based in scientific evidence. In fact, in a well-known blind study, it was the passive Revel Salon2 (which Floyd Toole uses as his personal speakers of choice at home by the way) that were much preferred over the active JBL M2, despite both measuring very well (possibly in slightly different ways though). So it's perfectly possible for passive speakers to be well-performing and designed based on evidence and science - a speaker should never be ruled out just because it's a passive design.

That people preferred the Ultima2 Salon over M2 is interesting because the former produces a shallow dip in the presence region (off-axis) whilst the latter does a shallower dip in the midrange (off-axis):

 

[q3cpma]

That people preferred the Ultima2 Salon over M2 is interesting because the former produces a shallow dip in the presence region (off-axis) whilst the latter does a shallower dip in the midrange (off-axis):

Weren't you the one talking about sample bias in another thread? The kind of people in this sample may very well be used to a wider directivity.

About the passive vs active question, passive crossovers are "ruled out" by many more because of their technological and price deficiencies than final audio quality; not that audio quality is the same, especially on 2-ways where cone breakup due to shallow crossover slopes matter.
Anyway, the M2 and Ultima2 are simply way too different to be used as crossover comparisons.

 

[TimVG]

If you were to look at the lateral data only (which the spinorama does not show) - you would see the Salon2 does not feature an off-axis dip and that dispersion itself remains very wide up about 10kHz.

 

[tuga]

If you were to look at the lateral data only (which the spinorama does not show) - you would see the Salon2 does not feature an off-axis dip and that dispersion itself remains very wide up about 10kHz.

You may be replying to someone else because I said nothing about crossovers.
Which makes you wonder why it is showing in the spinorama if it is not representative of the speaker's behaviour...

But the NRC measurements do show a dip and that dip appears to live in a narrow shelved up band. Weird stuff.

 

[Putter]

What seem to be lost in this thread (at least as far as I can recall in a 35 page thread) is real world performance acoustics. Without reviewing papers on this of which I'm sure there are many, if you read older reviews of speakers in mags like High Fidelity and Audio you'll often find speakers given as having a 'front row' presentation based on relatively elevated treble vs. a 'middle row' presentation with a reduced treble content.

I would say most people would rather sit in the middle row sonically speaking to get a more balanced presentation, i.e. a more spacious sound. This would seem to definitely apply to relatively live performances while I'd reserve judgement on studio based recordings, but would suspect it would follow the same pattern.

 

[napilopez]

That people preferred the Ultima2 Salon over M2 is interesting because the former produces a shallow dip in the presence region (off-axis) whilst the latter does a shallower dip in the midrange (off-axis):

If you were to look at the lateral data only (which the spinorama does not show) - you would see the Salon2 does not feature an off-axis dip and that dispersion itself remains very wide up about 10kHz.

I believe Dr Toole said that he believes the reason the Salon2 'won' this specific comparison was because of its wide horizontal directivity. Indeed, the Salon2 has some of the widest horizontal dispersion I've seen from a dome tweeter, moreover one with a waveguide.

This is the main thing I don't like about the spinorama: it makes it pretty hard to eyeball how 'wide' directivity is and makes little distinction between vertical and horizontal effects. You can sometimes see the impact by looking at the difference between the early reflections and sound power curve(since the early reflections curve includes only a handful of vertical curves but sound-power is the full shebang), but the Salon2 doesn't show much of a difference. Contour plots often use different color grades and scaling.

But good ol' SPL charts, make it clear what's happening at every angle. You can estimate directivity simply by looking at the SPL difference between the 0 curve or listening window and the 75 degree curve. In this case normalized plots can be handy too. In both cases it will also be evident by the spacing between the various angle curves.

You can see, for example, how much wider the Salon2's directivity(and the ultima line in general) is than some of its lower-tier siblings.

Performa M126Be (I believe all of the PerformaBe series use a similar waveguide and should have similar results). First Listening Window:

Now out to 75 degrees off axis:

There's a steep roll off after 5kHz. By 8kHz @75 degrees, you're down 10-12 dB.

Salon2:

It's immediately apparent the Salon2 has wider directivity. The curves remain tightly spaced, and by 8KHz, you're only down about 6-8dB from the Listening Window, plus the curve's overall shape mirrors the listening window more closely. The Salon2 will therefore sound 'bigger' and have a wider sweet spot on directivity alone.

That said, vertical reflections can still affect tonality. Though it's not something I think designers should aim for, it's possible that an early reflections dip in the presence region is less offensive than if it existed elsewhere.

(And for an example of a very high directivity speaker, the Klipsch R-820F, down about 20dB)

Lastly, it's rather important to reiterate that though it seems a majority people prefer wider directivity, it is not like linear frequency response where the trend is for basically everyone to prefer a more linear response, with some flexibility for age and tilt. Some people have strong preferences for different directivity characteristics, and that was evidenced in the Salon2 vs M2 Blind test as well. It's a point that is sometimes forgotten in this comparison.

"We had 11 critical listeners over the last two days, plus myself and my crew, for a total of 15 scores. Out of the 15, three people consistently preferred the M2s over the Salons2, while the rest preferred the Salon2s generally. We also had three people who consistently preferred the Salon2s down the line, track by track, while the "middle group" tended to prefer the Salon2s, tilting the results in their favor."

While the average listener preferred the Salon2, Three people consistently preferred the M2's, and three consistently preferred the Salon2.
It probably comes down to how much you value "focus" and "imaging" versus "spaciousness."

This might also suggest that in a different listening room - perhaps a very live one where there are already many early reflections - the results might be skewed differently. It's been noted people who people tend to prefer more attenuated reflections while mixing and louder reflections while mastering or listening recreationally.

So once again, when it comes to sound, it seems directivity characteristics are where individual preferences really come through.

You may be replying to someone else because I said nothing about crossovers.
Which makes you wonder why it is showing in the spinorama if it is not representative of the speaker's behaviour...

But the NRC measurements do show a dip and that dip appears to live in a narrow shelved up band. Weird stuff.

Note sure what you are referring to That 'dip' is hardly present in the NRC data. A shallow scoop of roughly 1-1.5dB. Audible but unlikely to be a significant 'tuning' choice. The dip in the early reflections and power curve of the spinorama is due to the vertical data.

 

[Rick Sykora]

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.

Lol. This looks like the Polk SDA rewarmed from the 80s!

See page 32 from Audio June 1984:

https://www.americanradiohistory.com/Archive-All-Audio/Archive-Audio/80s/Audio-1984-06.pdf

 

[napilopez]

Lol. This looks like the Polk SDA rewarmed from the 80s!

See page 32 from Audio June 1984:

https://www.americanradiohistory.com/Archive-All-Audio/Archive-Audio/80s/Audio-1984-06.pdf

Yeah, apparently they've refined the tech with the angled baffle and some HRTF related tuning. I have no idea how it would measure, but the non-SDA L200 performs decently.

 

[thewas]

Indeed, the Salon2 has some of the widest horizontal dispersion I've seen from a dome tweeter, moreover one with a waveguide.

What is often misunderstood is that a good waveguide not only narrows the radiation on the lower range of a tweeter, making it match better the directivity of the mid/woofer, but often also widens the radiation of the tweeter in high frequencies.

 

[mitchco]

While the average listener preferred the Salon2, Three people consistently preferred the M2's, and three consistently preferred the Salon2.
It probably comes down to how much you value "focus" and "imaging" versus "spaciousness."

So once again, when it comes to sound, it seems directivity characteristics are where individual preferences really come through.

Sorry for the big snip of your post, but I think your statement above hits the nail on the head.

I have listened to both of those speakers before and I prefer the M2 over the Salon2. I prefer the focused imaging over the spaciousness. I tried an experiment comparing a pair of JBL 4722 Cinema speakers versus KEF LS50 (both with subs to take out LF comparo part) and preferred the much narrower directivity of the JBL's which are narrower than the M2's:

The top directivity curve is the JBL 4722, middle M2 and bottom Salon2.

Back to the JBL cinemas versus the KEF LS50, I made binaural recordings of both in the above link, and you can really hear the direcivity differences between the two. More direct sound with the JBL's and a lot more room sound with the LS50's. I had quite a few PM's about which one sounded best, and it was a 50/50 split or close to it.

So preference seems to reign at this level. This assumes that both speakers have controlled or constant directivity. Personally, I cannot stand speakers with really rising directivity in the last two octaves that give the speaker the "head in vice" listening window. Not for me.

 

[Rick Sykora]

Yeah, apparently they've refined the tech with the angled baffle and some HRTF related tuning. I have no idea how it would measure, but the non-SDA L200 performs decently.

I recall liking the sound of the original design but the dealer was a boutique shop and only sold at list price. They also required a special cable between the stereo pair and was not fond of that. Over 30 years later, I bought my CBT24s and pretty sure they sound better. Will have to see if they are auditioning them around here.

I think the original SDA was over $1000 back then, so about $2500 now. Ofc, that was around when first IBM PCs were just becoming widely available and the US still had a significant manufacturing base. Guessing am giving away my age but still think am younger than Amir!

 

[HammerSandwich]

Some people have strong preferences for different directivity characteristics, and that was evidenced in the Salon2 vs M2 Blind test as well. It's a point that is sometimes forgotten in this comparison.

Good note.

Without evidence, I'll guess that narrower dispersion scores relatively lower in mono. Spaciousness via reflections has to be more important when you don't have stereo/surround, right?

 

[napilopez]

Good note.

Without evidence, I'll guess that narrower dispersion scores relatively lower in mono. Spaciousness via reflections has to be more important when you don't have stereo/surround, right?

Yes, this seems to reflects Dr Toole's general findings if I'm interpreting things right. In general, spaciousness is associated with better sound quality, and wider dispersion speakers tend to sound more spacious.

But stereo significantly muddies the picture (and more channels moreso). Rankings are overall the same - and this has been shown over several tests -but grouping is significantly closer in stereo. It's harder to tell apart sound quality in general in stereo. Here's one example from the Toole study I referenced earlier.

It's an older study, but as far as I know, there isn't much to invalidate it. In 1985 Toole compared three speakers in a blind test conducted in both mono and stereo. All three have similarly flat on-axis, but rather different off-axis patterns. Two of them, the the Rega Model 3 and Kef 105.2, were in the wider directivity camp but had uneven off axis. The Quad ESL 63 was narrower directivity but had more timbrally balanced off-axis (Toole said it was expected the Quad would win).

Here was their frequency response:

And these were the results in mono and stereo, for both sound quality (tonality, presumably), and spatial qualities.

This test should showed two things: sound quality is surprisingly correlated with spatial quality, and people had strong feelings about spatial quality even in mono. On the other hand stereo seemed to significantly mask the perceived flaws in mono, even if rankings were overall similar. The Quad was wayyyy behind in mono, but lost by a comparitovely small margin in stereo.

(All this from the latest edition of Toole's book Chapter 7.4.2)