Evidence-based Speaker Designs

[maty]

Forgive the scales.

 

[Cosmik]

In this article, the author says

What is true of the mini-monitor, that it cannot be EQed to sound right, is also true of narrow-front floor-standers. They sound too midrange-oriented because of the nature of the room sound. This is something about the geometry of the design. It cannot be substantially altered by crossover decisions and so on.

Does this constitute "evidence"? What I take from this kind of statement is an idea that supersedes all measurements. It says that there is a fundamental limitation on small speakers of conventional design and there is nothing you can do to change it.

Pretty graphs, whether in-room or in anechoic chambers will record the phenomenon, but then the interpretation of the graphs is just as subjective as sighted listening.

You can do all the listening tests you like, but there is no point: the mini-monitor or narrow-fronted floorstander will *always* sound wrong if listened to at a distance.

You may disagree with the idea, but it comes from logic backed up with other ideas and references to science, as well as subjective experience. Without such ideas, no progress would be made; there wouldn't even be any speakers to measure. Measurements and listening tests are 'passive' evidence; you can't create something from evidence alone.

 

[Ilkless]

In this article, the author says

Does this constitute "evidence"? What I take from this kind of statement is an idea that supersedes all measurements. It says that there is a fundamental limitation on small speakers of conventional design and there is nothing you can do to change it.

Pretty graphs, whether in-room or in anechoic chambers will record the phenomenon, but then the interpretation of the graphs is just as subjective as sighted listening.

You can do all the listening tests you like, but there is no point: the mini-monitor or narrow-fronted floorstander will *always* sound wrong if listened to at a distance.

You may disagree with the idea, but it comes from logic backed up with other ideas and references to science, as well as subjective experience. Without such ideas, no progress would be made; there wouldn't even be any speakers to measure. Measurements and listening tests are 'passive' evidence; you can't create something from evidence alone.

He seems to be talking about baffle-step compensation (or the lack thereof in many small narrow baffle speakers because it incurs more parts cost and reduces sensitivity), causing a thin balance. Baffle step and BSC are uncontroversial, quantifiable and predictable. The thing is when people like him couch it in arbitrary touchy-feely terms and claim, from a position of ignorance, that the phenomena elude empirical observation and can't be captured by our knowledge of acoustics.

 

[Cosmik]

He seems to be talking about baffle-step compensation (or the lack thereof in many small narrow baffle speakers because it incurs more parts cost and reduces sensitivity), causing a thin balance. Baffle step and BSC are uncontroversial, quantifiable and predictable. The thing is when people like him couch it in arbitrary touchy-feely terms and claim, from a position of ignorance, that the phenomena elude empirical observation and can't be captured by our knowledge of acoustics.

Yes, he is talking about baffle step compensation being a compromise between neutral on-axis response and neutral off-axis - that it is not possible to make the speaker 'correct'.

This is something about the geometry of the design. It cannot be substantially altered by crossover decisions and so on

He is not unaware of the concept.

In a single sentence, he sums up why a small speaker sounds small. No measurements; no listening tests; effectively no empirical "evidence". What he says is logically correct, and armed with this idea, a speaker designer can do in an afternoon what the "evidence-based" designer would never achieve in his entire lifetime.

If you were a would-be speaker designer, and you saw that the hi-fi shops and shows were full of narrow-fronted speakers being touted as no-compromise transducers costing more than a new listening room, and that these were the only speakers being tested in listening tests, why on earth would you do anything different, yourself? That would be the only "evidence" available. Only the person armed with the idea can produce a better speaker. And ideas are not science, nor "evidence".

 

[Ilkless]

Yes, he is talking about baffle step compensation being a compromise between neutral on-axis response and neutral off-axis - that it is not possible to make the speaker 'correct'.

He is not unaware of the concept.

In a single sentence, he sums up why a small speaker sounds small. No measurements; no listening tests; effectively no empirical "evidence". What he says is logically correct, and armed with this idea, a speaker designer can do in an afternoon what the "evidence-based" designer would never achieve in his entire lifetime.

If you were a would-be speaker designer, and you saw that the hi-fi shops and shows were full of narrow-fronted speakers being touted as no-compromise transducers costing more than a new listening room, and that these were the only speakers being tested in listening tests, why on earth would you do anything different, yourself? That would be the only "evidence" available. Only the person armed with the idea can produce a better speaker. And ideas are not science, nor "evidence".

Are you aware of what is baffle step? Your intuitive understanding of it sounds as suspect as his. The Salk article is a good plain-English treatment of the issue. Baffle step is the transition region in which the wavelength of emitted sound gradually becomes large relative to baffle dimensions, causing sound to "wrap" around it. This causes SPL losses relative to frequencies emitted forwards (180°), where wavelengths are small relative to baffle dimensions and no such "wrapping" occurs. The losses max out at about 6dB - half the SPL (recall transition from 180° to 360°). Baffle step is crucial for optimal balance at any frontal axis, at a listening distance that is large relative to the baffle dimensions (particularly width).

On the point about hi-fi shows, why would any competent engineer rely on anecdotal observation when there is the wave equation and tons of research literature on measured SPL response of baffle step plus diffraction? Linkwitz provides another excellent treatment of both these concepts. I certainly would not trust any speaker designer not even acquainted with basic acoustics.

 

[andreasmaaan]

In this article, the author says
Does this constitute "evidence"? What I take from this kind of statement is an idea that supersedes all measurements. It says that there is a fundamental limitation on small speakers of conventional design and there is nothing you can do to change it.

In that article, the author states (in reference to an 18"-wide cabinet box speaker):

"Of course, at first sight, it might not seem all that advantageous to have the “baffle step” , the transition from omni to forward[“4 π to 2 π”] radiation, at the lower frequency here compared to the higher frequency of the minimonitor.

But a basic psychoacoustic fact already noted makes the difference: Up to around 300 Hz, the ear/brain hears balance as a unit, with no substantive distinction between direct sound, reflected sound, and room sound. The response heard is the total room response. But from 300 Hz on up, there is a transition into hearing the reflected sound and room sound separately from the direct arrival. What this means is that a speaker like that in example 3 can in fact be made neutral-sounding by DSP."

Whether or not his view is factually accurate, isn't this exactly what evidence-based speaker design is about?

 

[Cosmik]

Whether or not his view is factually accurate, isn't this exactly what evidence-based speaker design is about?

I am making a distinction between what is being worked out in his head versus the idea of driving a design from "evidence".

The latter suggests that you 'try stuff out' e.g. try building some narrow-fronted speakers and, if you are not simply copying what's on sale in the hi-fi shop, some wider monkey coffins. Based on what the evidence shows (in listening tests or measurements?), you pursue a particular avenue. At no time in this process is it necessary to know why the evidence shows what it shows, and this is where the process heads off into dead ends and (in neural network terms) 'local minima'.

I like the idea of someone like Mr. Greene identifying a mistake in the speaker design orthodoxy that is persisting because everyone is copying everyone else and not thinking it through - and possibly not having very good critical faculties to boot. He doesn't need to perform any experiments or publish any scientific papers to prove it. This is how to clear the fog: not by examining existing evidence, or even running one's own experiments - because these will get bogged down in experimental errors and misunderstandings.

The Grimm LS1 people had the same idea, too. They were never going to build a narrow speaker once they had identified the problem: they weren't interested in scientific experiments; they were just going to build a wide speaker.

 

[andreasmaaan]

I am making a distinction between what is being worked out in his head versus the idea of driving a design from "evidence".

The latter suggests that you 'try stuff out' e.g. try building some narrow-fronted speakers and, if you are not simply copying what's on sale in the hi-fi shop, some wider monkey coffins. Based on what the evidence shows (in listening tests or measurements?), you pursue a particular avenue. At no time in this process is it necessary to know why the evidence shows what it shows, and this is where the process heads off into dead ends and (in neural network terms) 'local minima'.

I like the idea of someone like Mr. Greene identifying a mistake in the speaker design orthodoxy that is persisting because everyone is copying everyone else and not thinking it through - and possibly not having very good critical faculties to boot. He doesn't need to perform any experiments or publish any scientific papers to prove it. This is how to clear the fog: not by examining existing evidence, or even running one's own experiments - because these will get bogged down in experimental errors and misunderstandings.

The Grimm LS1 people had the same idea, too. They were never going to build a narrow speaker once they had identified the problem: they weren't interested in scientific experiments; they were just going to build a wide speaker.

I of course agree with you that logic needs to play a role. But how did the Grimm people decide how wide "wide" was? Their baffle is (from recollection) about 50cm wide. That's nowhere near wide enough to control the directivity of the lowest frequency the speaker reproduces. How did they decide what was "enough"? Did they just say "that's all the wood we can afford"?

As I've said before, the best speakers combine theory and experimental evidence.

 

[Cosmik]

I of course agree with you that logic needs to play a role. But how did the Grimm people decide how wide "wide" was? Their baffle is (from recollection) about 50cm wide. That's nowhere near wide enough to control the directivity of the lowest frequency the speaker reproduces. How did they decide what was "enough"? Did they just say "that's all the wood we can afford"?

As I've said before, the best speakers combine theory and experimental evidence.

Yes, I'm sure there's an element of chicken and egg, but in the corridors of ASR, a great deal of attention is placed on retrospective "evidence" e.g. a huge database of speaker measurements was being advertised for $8.90 in a thread this morning - I'll bet that most of them resemble each other. And people love their listening tests and statistics.

Now, if you are Mr. Greene you don't need to buy this database and pore over the graphs! If you know that the speaker is going to be used in your living room and it's narrow-fronted, you can dismiss it out of hand, because you *know* it can never be made to sound right. No evidence (except dimensions measured with a tape measure) is necessary.

 

[Ron Texas]

In this article, the author says

Does this constitute "evidence"? What I take from this kind of statement is an idea that supersedes all measurements. It says that there is a fundamental limitation on small speakers of conventional design and there is nothing you can do to change it.

You can do all the listening tests you like, but there is no point: the mini-monitor or narrow-fronted floorstander will *always* sound wrong if listened to at a distance.

Would someone who is familiar with the work of @Floyd Toole comment on the above statement and article link. By the way, what is "a distance"? Is it further than a desktop, further than 8', or real far, like in an auditorium?

What I do know is most speakers today are relatively narrow and some are rounded off at the top to make them narrower. Anyway, some further explanation is necessary because this seems at odds with other parts of the discussion where a preference for wide dispersion was discussed.

 

[HammerSandwich]

But how did the Grimm people decide how wide "wide" was? Their baffle is (from recollection) about 50cm wide. That's nowhere near wide enough to control the directivity of the lowest frequency the speaker reproduces. How did they decide what was "enough"? Did they just say "that's all the wood we can afford"?

See Grimm's "What were we thinking?" PDF, which cites the same ~300Hz transition as REG. Beyond that, it's an excellent primer on how to engineer a modern, DSP speaker.

 

[Juhazi]

Juhazi said:
Horizontally a CBT is nothing special, but room response is actually 3D directivity plus time including reflections and modes. CBT helps helps a lot with reflections.
---
That's not true.

First of all, one can't make a general statement about the horizontal directivity of a CBT. Why? Because it can be implemented in various ways. Both with various kinds of directivity and with different types of shading networks that will effect the performance.

If you take the first CBT on the market, CBT36, it has a constant and wide horizontal directivity down to approximately 400 Hz. While it's not perfect and has a few side lobes, it's still very uniform and certainly way above most speakers out there. You can see that clearly illustrated in Don's video where he measures it one the sides in real time. Look after 6 minutes in video part 8.

This is cropped from Keele's CBT36 vs. B&W801 paper http://www.audioartistry.com/products_CBT.htm

I can't say it is constant or wide. And how about another direction of rotation? One might say "constant up to 400Hz", but that is common to nearly all speakers, specially in small rooms with long gating, including modes and reflections. In the video there is RTA measurement with ARTA.

 

[Bjorn]

IMO it's a shame so few publish vertical polars. But there's a good reason they don't.

This is cropped from Keele's CBT36 vs. B&W801 paper http://www.audioartistry.com/products_CBT.htm

I can't say it is constant or wide. And how about another direction of rotation? One might say "constant up to 400Hz", but that is common to nearly all speakers, specially in small rooms with long gating, including modes and reflections. In the video there is RTA measurement with ARTA.

View attachment 22364

It's definitely very uniform in the horizontal as well. Also wide (180 degrees), which is evident from the polars. Take note that the measurement above isn't truly anechoic, it's with an uneven floor (may or may not matter), uses 1/12 smoothing and the decadent (resolution of graph) is 25 dB. With a graph resolution most show and with 1/3 smoothing it would look a lot better. But sure, the CBT36 has some side lobes.

I said constant down to 400 Hz. Not up. Two different things.

 

[Juhazi]

Major audio companies that have the leading edge with research and development of loudspeakers are
1) Harman group, speakers JBL, Revel Have had many of the leading electric and acoustic designers and have published lots of research, excellent web pages, long history starting from Lansing
2) KEF, innovators and developers of coaxial drivers and speakers, Uni-Q with tangerine phase plug
3) Bang & Olufsen BeoLab speakers, many unique constructions, modern flagship BeoLab 90
4) Genelec, many innovations and very high and consistent quality and value in pro audio (aluminium cases, waveguides, coaxials)
( 5) British speakers during 60-70s, starting from BBC-guidelines for monitoring, Rogers, Harbeth)

and then many other large manufacturers like Sony, Technics, Bowers&Wilkins, Elac etc. with some nice flagship or spot-on models
and then lots of small companies run by inventors, GedLee, Gradient, Vivid Audio, Magico, YG, etc.
And then amateur designers with diy projects Linkwitz, Kreskowsky, Keele, "Marsh", Krutke, Gravesen etc.

 

[Krunok]

In this article, the author says

Does this constitute "evidence"? What I take from this kind of statement is an idea that supersedes all measurements. It says that there is a fundamental limitation on small speakers of conventional design and there is nothing you can do to change it.

I don't see he offers much proof (of evidence) for that idea.

In this article, the author says
You can do all the listening tests you like, but there is no point: the mini-monitor or narrow-fronted floorstander will *always* sound wrong if listened to at a distance.

If you, when you say distance, mean 4-6m I can agree with the mini monitor part but certainly not with narrow-fronted floorstander.

 

[Ilkless]

Major audio companies that have the leading edge with research and development of loudspeakers are
1) Harman group, speakers JBL, Revel Have had many of the leading electric and acoustic designers and have published lots of research, excellent web pages, long history starting from Lansing
2) KEF, innovators and developers of coaxial drivers and speakers, Uni-Q with tangerine phase plug
3) Bang & Olufsen BeoLab speakers, many unique constructions, modern flagship BeoLab 90
4) Genelec, many innovations and very high and consistent quality and value in pro audio (aluminium cases, waveguides, coaxials)
( 5) British speakers during 60-70s, starting from BBC-guidelines for monitoring, Rogers, Harbeth)

and then many other large manufacturers like Sony, Technics, Bowers&Wilkins, Elac etc. with some nice flagship or spot-on models
and then lots of small companies run by inventors, GedLee, Gradient, Vivid Audio, YG, etc.
And then amateur designers with diy projects Linkwitz, Kreskowsky, Keele, "Marsh", Krutke, Gravesen etc.

Agree with pretty much everything on your list, Rogers and Harbeth were formidable for their time, though they have barely moved on since.

 

[Krunok]

Few months ago I have tried this experiment related to difference between mini monitors vs floorstanders. As I have these little guys on hand acting as surrounds I tried to compare them vs my front floorstanders. Interesting thing is they have the same drivers. Difference is of course that Castle Harlech S2 have 2 bass drivers and they are transmission line while Richmonds are sealed boxes.



My listening position is app 4m away from each speaker. As I don't have stands I have put Richmonds on top of Harlechs. The only difference I was pretty much expecting was in bass response but that turned out to be completely wrong. Even when pushed to their limits Richmonds didn't manage to fill the room (app 40m2) with energy and they sounded what I can only describe as "hollow". In a smaller room and when listened from shorter distance they actually sound very well.

 

[andreasmaaan]

Would someone who is familiar with the work of @Floyd Toole comment on the above statement and article link.

Greene’s statement doesn’t directly accord with my understanding of Toole’s research. It’s been a while since I read Toole, but IIRC the findings were that speakers with a flat axial response and constant directivity or downward-sloping power response were preferred. There was no specific finding regarding preferred baffle width to my knowledge.

Revel speakers are all “narrow”-baffle too FWIW.

 

[andreasmaaan]

It's definitely very uniform in the horizontal as well. Also wide (180 degrees), which is evident from the polars.

I look at that and see a wide horizontal polar response, but hardly a uniform one when compared to a speaker with conventionally-arrayed drivers and in which attention has been paid to directivity matching in the design.

Compared to Kii Three for example:

Without a doubt, the CBT36's vertical polars are miles ahead of any conventional design of course.

 

[maty]

The two philosophies in a picture.

[Review] Revel Performa3 F206 and vs Harbeth Monitor 40.1

http://highfidelity.pl/@main-405&lang=en

https://revelspeakers.com/productdetail/~/product/f206.html

https://www.harbeth.co.uk/hifi-speaker/20/Harbeth-Monitor-40.2-loudspeaker