Evidence-based Speaker Designs

[Kal Rubinson]

A few weeks ago the rage here was DSP controlled active speakers. Now it's large baffles. What will it be in March?

March comes in like a lion. Is that a clue?

 

[SIY]

The cabinet is 50 cm wide. Making it wider would move the baffle step lower in frequency, but I felt 50 cm was as wide as I could go and still being able to reach a market. Few can live with a very wide speaker and that's why most speakers are made narrow.

So, in-wall would be optimum?

 

[RayDunzl]

My consistent impression of the Harbeth sound (in the SuperHL5plus and up range) is one that sounds *particularly* rich and full. Vs thin and anemic or reductive.

My recent very brief exposure to those Harbeths at the Florida Audio Expo left me wanting to hear more.

I'm immune to high frequency, and like my big old antique MartinLogans, but the Harbeth setup made me take notice. Only two or three setups at the show did that for me, and, perhaps, the Harbeth was the most surprising, being an "old school" box with cones and domes.

I get to be wrong, one tune, five minutes exposure, maybe. I think it was Dean Martin, of all people. My Audio Buddy later commented something like "I never realized Dean Martin's vocalizations had so much ____________ (I forget, exactly)"

https://www.stereophile.com/content/opening-ears-wider-day-one

J.V.S: "A beautiful warm midrange was the standout feature of a system that featured Luxman electronics and Harbeth loudspeakers."

I thought it something more than just that. Some sort of room-filling hugeness yet tightly focused. I like big. I like focus.

*cranks up Spotify with some Dean Martin

 

[Juhazi]

Most modern floorstanders' baffle width is around 180 - 300mm, which is a very small difference regarding baffle step. 500mm starts to be wider (Grimm LS-1), but around 1000mm we are in really different territory, specially if we use a curved baffle (SF Stradivari)

GoldenEar Technology Triton is 172mm wide, but has several bass drivers and side-mounted passive reflex units. Testers didn' call it bass shy.

Conclusion - Baffle width is not dictating sound, it is more about how directivity above baffle step develops and how much of baffle loss is handled (read = power response profile).

 

[Bjorn]

So, in-wall would be optimum?

They have definite advantages and are common in studios. But they have one strong weakness IMO; the depth in sound stage you loose, something that's probably mostly a visual psychoacoustic effect.

Whether a speaker sounds "thin" or "full" also depends much on how it's tuned and how it integrates in the room.

 

[MattHooper]

It is hard to accept as reliable this un-blinded comparison given the visual influence on the listener. Also, even if one accepts the described audible dichotomy between the two speakers, they afford no relationship to accuracy, just personal preferences as revealed by the chosen adjectives.

Yes, which is why I qualified my observations. And I also wasn't speaking to accuracy, only the apparent subjective effects.

Though this brings to mind the problem we enter when any reports, or even one's own experience listening to speakers, are dismissed if not done double-blind.

It seems to be one thing when talking about areas where the very technical claims are dubious - e.g. those behind any number of audiophile tweakdom like power cables. I would not trust anyone's anecdote in areas where the very audibility of a phenomenon is in dispute by people educated in the subject (e.g. EEs etc).

But it becomes a question of how far to take this. We don't normally reject one another's, or our own, impressions where there are good empirical reasons to think differences are discernible - e.g. "are you SURE your orange juice actually tasted different than your milk? I'm not going to accept this unless you've done double-blind experiments." Yes, bias is always there, but it's also the case that there are objective reasons for why orange juice would taste discernably different from milk. Surely we would end up being crippled by such stringent demands on our experience and can provisionally accept claims of differences in every day life, where differences have some plausible basis.

So where do speakers fall in to this equation? Speakers are acknowledged to have objective differences in performance that correlate to detectable sonic differences. At least on the surface, it seems reasonable to relax the stringent demands somewhat on personal experience. In other words, if someone thinks X speaker sounded different from Y speaker....that's entirely plausible that they sounded different...unlike technically controversial claims for cables etc.

And then that person can reasonably say "I preferred X over Y."

But then it has to be acknowledge when we are getting really careful about our conclusions, that bias operates even at the level of deciding between items that do, in fact, sound sonically different! Which is the point you make, and the point made by research like Floyd Toole and HK's facilities.

So we don't escape bias effects even when comparing audio components like speakers that do have detectable sonic differences.

But, again, how far do we reasonably take the demands for blind testing? If we want to be sure of a claim to a scientifically acceptable degree....blind testing is a great tool. But how reasonable would it be to therefore dismiss any worth of an individual's experience comparing two speakers outside the lab? If we dismiss it all, it seems to me going too far. It's true that when I compare, say, a smaller Harbeth box speaker with big Avante Guarde horns, that some level of visual/conceptual bias may be playing part of my impressions. But it is also plausibly the case that the speakers do sound quite different, and in the ways I may detect and describe.

If we want to be so rigorous in assessing speakers that we simply can not rely on our own reactions and impressions to a speaker, that seems to be rather severe. It would seem we would be stuck only considering those speakers from companies - quite rare at the moment - providing spinorama-like measured documentation of their products. And so HK can tell us "trust us...we've done the measurements of our products vs competitors...you'll prefer our products, no need to bother listening to the competitor as we've done the technical work."

Personally, I don't like narrowing the field that much. I've encountered so many different types of speakers that I've really enjoyed, that would not necessarily come out on top in HK's blind testing facilities. And when I've auditioned the Revel speakers, I've found them very well designed and competent, but did not love the sound as I have any number of other speakers. So do I simply dismiss my own experience and say "even though I've listened to the Revel speakers and found myself not moved by the sound, I should just buy Revel anyway because they've done tests suggesting most people would prefer them under blind conditions?"

The blind tests give us statistical probabilities. On HK's work, I'm statistically likely to prefer the HK speakers over, say, the Devore speakers I *think* I enjoyed more. But then, I may fall within the statistics of those who wouldn't pick the Revels. I can't know unless I take a trip to Revel and do a blind shoot out between Revel and any number of other speakers I may be interested in. But that's frankly not a workable idea for the vast majority of us.

So my point being: I absolutely agree that if we want a scientific level of confidence about the type of impressions I've described...or simply in terms of comparing speakers in the consumer context, then we have to acknowledge the factor of bias and control for it as HK does.
But in terms of every day practicality, it seems reasonable to be less stringent and be able to talk with one another about what he heard from one speaker to another. Otherwise it would seem all intersubjective references and discussions about how speakers sound is to be dismissed or banned as inconsequential and disallowed. "So you bought new speakers? How do they sound compared to your old ones?" "Sorry, can't have that conversation. I don't have access to a testing facility where I can do double blind comparisons of speakers." Which sounds like a world that really sucks to my mind.

Yes, I'm sure very few would take things to the extreme I'm envisioning...which is my point. Objective data is obviously very important. But there can start to be a sort of tyranny-of-measurements to some degree, where one's own input, or even technical claims, is never good enough as it can always be challenged on technical grounds. I believe you started to experience a bit of that on the AVSforum for your still enjoying and lauding of B&W speakers even though you are aware of the technical/scientific case made for Revel speakers. "How can you say those speakers sound good - they don't measure great and excell in double blind tests like Revel!"

 

[Kal Rubinson]

Yes, which is why I qualified my observations. And I also wasn't speaking to accuracy, only the apparent subjective effects.

I was sorta underscoring the point because you were making a specific correlation of the sound with the appearance (width).

Though this brings to mind the problem we enter when any reports, or even one's own experience listening to speakers, are dismissed if not done double-blind.

Again, there are obvious conceptual/linguistic/logical correlations of wide with "rich and full" vs. narrow with "thin and anemic....."

So, I am not really arguing with your preference or, indeed, with your perceptions but the opportunity for subconscious bias seems obvious in this particular case. I know that I am vulnerable to the same biases as everyone else. OTOH, if I wanted to draw a general conclusion about a sonic relationship between some sound quality parameter and a visual/physical parameter, I'd want the obvious associations obscured.

 

[MattHooper]

Again, there is an obvious logical correlation of wide with "rich and full" vs. narrow with "thin and anemic....."
So, I am not really arguing with your preference or, indeed, with your perceptions but the opportunity for subconscious bias seems obvious in this particular case.

Absolutely, I agree.

Though I think the concerns I raised, in terms of what are we to allow from our subjective evaluations, still follow from this problem.

BTW, whenever I audition a speaker, for this reason I always do several "close my eyes" tests to notice the character of the sound when I'm not directly looking at the speaker. I ask myself "if I didn't know what speaker I was listening to, and just concentrate on the sound, what does it sound like? How big, small etc would I guess this speaker is?"

Now, very obviously that is nothing near actual blind testing in terms of addressing bias. But, lacking facilities to blind test, I have still found it an interesting test, because sometimes the sound with eyes closed really does not match the appearance of the speaker. I've auditioned speakers that were quite large and yet when I closed my eyes and evaluated how big the sound really seemed, I'd have guessed a far smaller speaker.
I remember first starting this when I was long ago listening to my smaller pair of thiel 02 monitors, which came with small stands holding them only inches above the floor, angled upward. Eyes open, seeing the placement of the speakers, the soundstage tended to hover closer to the floor. But I could close my eyes, and simply picture that I was listening to tall floor standing speakers and it was almost miraculous as the sound gradually "lifted" in my perception upward! Open my eyes, and down it went. Reminds me also of watching movies in the drive in through the car stereo system. When watching the screen it was always amazing to me how my brain mapped the sound right on to the far away screen in view.
But if I closed my eyes, the visual influence gone, the sound went right to the speakers in my car.

Again, very far from the usefulness of blind testing, but the closed eyes test does seem to yield some interesting results. In the close-my-eyes test the Devore speakers continued to yeild very big, fat sonic images. I had played a particular drum solo track on many different speakers using the "close my eyes, could it be a real drum set?" test, and the sensation of being in the presence of a drum set being played in front of me was more convincing on the Devores vs a number of other speakers. Again...lacking my own blind testing facilities...I gather my own impressions the best I can.

 

[Cosmik]

I think some of the above posts back up my point: any particular implementation of a speaker may, or may not, be well done in all sorts of ways. It is certainly possible that a physically large speaker will sound 'small', and that a physically small speaker sounds larger than it. This would be the problem with developing a speaker based on "evidence": the evidence is contaminated with noise.

At the end of performing listening tests on 100 existing speakers you would still have no clear idea of the role of baffle width on the sound - how were those speakers EQ'ed?; where were they placed in the room?; how far was the listener? etc. Had the designer already responded to the sound by voicing it to try to sound like another speaker? etc.

Even if you tried a more controlled experiment by changing only one variable - the baffle width - you still wouldn't know that in doing so you hadn't had the speaker set up wrongly all along or whether controlling just that single variable was, in fact, a meaningful test (maybe you need to change the baffle width *and* the EQ at the same time).

Only the R.E.G./Grimm approach can cut through this noise. They gain a theoretical understanding of the role of baffle width and then stick with it, ignoring spurious "evidence". It is one less variable to play with.

 

[DDF]

Yes, but I don't really know what a "power hole" is.

Been designing speakers (scientifically) for 40 years and IMO most good designs aim for a nominally flat on axis, controlled off axis and then tweak on axis if the off axis is not so well behaved. There is allot of vague psycho-acoustic data out there on how room sound affects tonal balance vs direct sound (direction of reflection and delay from incident being principle factors in whether tonal balance is affected) but to a first degree, designs tuned for nominally flat on axis will have a different tonal balance if they have different directivity (as we all know). The large baffle changes the directivity. I`ve attached my study of the Grimm to save anyone a trip to DIYaudio and made some conclusions therein about possible tonal impacts.

Large baffles don`t sound ``big`` except from the perspective of headroom and bass depth if designed correctly.

We should also discuss how the edge diffraction effects change with larger baffles. Large delays due to edge diffraction smear the image more but have less impact on tonal balance if the delay gets greater than about 5 ms (Ken Kantor`s magic speaker discussed this in detail)

 

[Juhazi]

Thank you David! I had the feeling that I have read that analysis earlier somewhere in the internet... Like mentioned earlier, driver choices and crossover must match and support or counteract the baffle effect. Balanced power response and smooth increase of directivity index are key elements.

I mentioned another wide baffle speaker previously, I guess your text applies to it as well. Sonus Faber Stradivari's baffle is 20" and curved. I would like to diy a copy, but I live in a too small house... A clip from Stereophile: the Stradivari presented a more weighty, unusually solid picture that seemed to be a three-dimensional curtain wrapped behind the baffles and extending well back into virtual space.
Read more at https://www.stereophile.com/content...homage-loudspeaker-page-3#g8M7wYchDdTvazVi.99

And here is a diy project with measurements KS-483 and comment about sound "Listening range is very wide. Tonal balance is surprisingly neutral and bright enough even 90° off-axis. Speaker's front wall acts like wide and flat wave guide, preventing diffractions and sharp turn-down in off-axis responses around 1 kHz - like many regular 2/3-way box speakers do. "

 

[Cosmik]

Large baffles don`t sound ``big`` except from the perspective of headroom and bass depth if designed correctly.

But your study says,

The response differences off axis relative to on axis are pushed further below the critical midrange, and may be less audible and less bothersome

This is the point being made by R.E.G.: that the (well-designed) narrow monitor is "too midrange oriented" and cannot be EQ'ed to sound the same as the (well-designed) wider speaker box because it is an issue of directivity (which includes the interaction with the room in the time domain, too).

This may be indicated by the relationship between on-axis frequency response and bucket-based "power response". (But this is not the same as saying that the relationship between those two measures is the objective of the exercise; it is an indication of an underlying characteristic that is more usefully defined in theory).

If this could be true for any large speaker but not a small one, then surely there is a "small" sound and a "big" sound..? (where "big" could mean 'more neutral' or 'different' rather than "big" per se).

The evidence of the "power response" curves shows this phenomenon indirectly and they need to be interpreted subjectively. Without the Grimm LS1 paper to direct you, would you have interpreted the graphs in the same way, or run the study in the same way?

 

[Shadrach]

My personal theory is that it has to do with the fundamental limitation of two-channel stereo, in which some of the sound almost inevitably emerges from the L or R loudspeakers - hard panned by console controls or by mic techniques. When this happens close miked solo instruments or groups of orchestra instruments emerge from a point in space. I have found this to be highly annoying in some recordings as it is plainly unrealistic. Spraying the sound around adjacent walls softens the mono L&R images. It also softens the amplitude panned images between the loudspeakers which some people disparage and others find appealing. The musical genre is likely a factor. But the basic issue is that conventional stereo cannot deliver the goods and we look for ways to improve it.

This is pretty much my view of stereo sound reproduction.
I'm very fond of classical music but I can't listen to it on a stereo system; there is just too much information missing, or wrongly placed to make the experience enjoyable. The idea that you can have fidelity to a live classical concert while a goal worth pursuing perhaps, hasn't been reached by the conventional two channel system, nor by any multi channel system I've heard.
There is I think a fundamental error in the views of many that what they listen to on their stereo systems, or what they hope to hear from their systems is music; it isn't, it's a selection of signals which may, or may not produce the illusion of recorded music.
I'm content to listen to most other genres; mostly electrified and my personal debate then is, do I want to have high fidelity to the source, or an enjoyable experience. The two may not be mutually exclusive.

 

[Juhazi]

About wide baffles, it might be that they mitigate SBIR cancellation coming form front wall, thus giving smoother bass response at 100-200Hz. A wide baffe is however omnipole at those frequencies, but still.. I haven't seen measurement to show this effect, JA sometimes does and publishes those, but not for Stradivari.

 

[Cosmik]

The way I am looking at it, a measurement may be evidence of something, but it is not (necessarily) 'the something' itself.

In economics, it may be claimed that an inflation level that lies inside some specified target range indicates a healthy economy, but it does not then follow that forcing the inflation rate to lie within the target range makes the economy healthy (probably the opposite, as demonstrated by former British chancellor of the exchequer Gordon Brown!). The validity of the assumption is based on the underlying economy being a self-regulating economy in the first place. By targeting the inflation measure directly, the nature of the economy is changed, and the 'target' loses its validity.

In audio, frequency response measurements may be an indication of something, but they are not the function of the system. In thinking that frequency response, or even sophisticated developments of it such as "power response" versus on-axis response, are the objective of an audio system, people are tempted to modify the system to force the frequency response to fit their 'target', like running a command economy that sets prices directly. Depending on how radical their interventions, they may kill the underlying 'purity' of the system.

For example, it might be that taking the back off a speaker (i.e. creating an open baffle speaker) 'improves' the frequency response curves, or that adding extra drivers that spray delayed, inverted audio around do something similar. But this risks adding entirely new time-domain dimensions to the sound that have got nothing to do with the recording, and effectively invalidate the validity of the frequency response measurements.

It is best to start with a metaphorical 'ideology' that works logically, and to stick with it with minimal interventions. For example, monopole-only, sealed box, wide baffle, single driver per way, etc. The Grimm paper makes this case, I think.

In politics there is the notion of "evidence-based policy" and this is held by many to be superior to "ideology", but I am not convinced. I think exactly the same thing about speaker design!

 

[Ilkless]

In politics there is the notion of "evidence-based policy" and this is held by many to be superior to "ideology", but I am not convinced. I think exactly the same thing about speaker design!

This reads to me like a devolution back to the very thing that ASR was founded to question - unfettered subjectivity of sound reproduction, except now couched in sophistry.

 

[Cosmik]

This reads to me like a devolution back to the very thing that ASR was founded to question - unfettered subjectivity of sound reproduction, except now couched in sophistry.

You have it the wrong way round. I am not recommending listening to speakers as a way of designing them or rating them. And I go even further: I don't even want to measure them! (except for confirmation they're working as designed).

I am saying that interpretation of a subset of "evidence" is no less subjective than a Michael Fremer review. I am saying that you should ideally know what is going to happen before you even build the speaker, and you should constrain its design in certain ways that guarantee a 'pure' response even if that means that individual aspects of its performance are not amenable to change or control. A Frankenstein speaker can be continually tinkered with in order to patch together 'better' measurements, but in reality it is ceasing to be a speaker and becoming a highly specific signal generator targeted to producing just those measurements. From the snapshot of "evidence" that results, you cannot tell that as the listener, if you move your head the dynamic phasing effects make you feel nauseous! (for example)

 

[Cosmik]

Putting it another way: the ideal speaker might produce a certain combination of on-axis and room response.

But coaxing your speaker to duplicate those apparent measurements (inherently ignoring 'the fur' that is a consequence of time domain behaviour your ears may not ignore; inherently not registering what the speaker does in other locations) does not mean it will necessarily sound like the ideal speaker.

It seems to me that this point eludes many people because frequency response is viewed as the objective of the speaker, not a by-product.

 

[RayDunzl]

except for confirmation they're working as designed

If they are working, how could they not be "working as designed" (even though the working result might be a surprise to the designer)?

Carry on!

 

[Mad_Economist]

In economics, it may be claimed that an inflation level that lies inside some specified target range indicates a healthy economy, but it does not then follow that forcing the inflation rate to lie within the target range makes the economy healthy (probably the opposite, as demonstrated by former British chancellor of the exchequer Gordon Brown!). The validity of the assumption is based on the underlying economy being a self-regulating economy in the first place. By targeting the inflation measure directly, the nature of the economy is changed, and the 'target' loses its validity.

This is a mangling of both inflation targeting and the Lucas critique.

The concept of ideas over evidence is a popular one in popular discussion of politics, economics, and audio - to give an economic example, since we're apparently doing that now, there still exist many who assert that a non-disemploying minimum wage is impossible or highly unlikely on a pure theory basis. Meanwhile, the people who have actually empirically studied the impacts of minimum wages on employment have advanced the theory in new ways to explain the actual results they observe, as science is supposed to.

Empirical work is core to the corrective feedback of science.