• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Why don't all speaker manufacturers design for flat on-axis and smooth off-axis?

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
Because it is not what you think. Nor is it relevant to the topic at hand.

The study I quoted goes to the precise need we as audiophiles have: which speaker is liked better, by a statistically significant number of listeners and what metrics predict that.

The study you linked to is not that at all. They set out to define a few subjective terms (e.g. BassPunch) and then try to see what measurements can correlate with what the rating of the trained listeners were in that regard. It is like trying to define some flavor in wine and seeing if a chemical make up can be correlated to it. It says nothing about whether said wine will be preferred by majority of people.

Indeed in the key metric of "natural," the study failed to generate useful results:

View attachment 29714

In addition as you see, they realized that these characteristics that were defined in advance were actually correlated in some cases. The ones they did use are all correlated to frequency response anyway.

The study is also very small scale and authors readily state that it cannot be generalized. Harman study is very extensive across many listening tests, and countless speakers, type of listeners, etc. This is why it has the weight that it has.

Bottom line: this study is orthogonal to what makes a good sounding speaker. They are trying to give you vocabulary to describe the sound of a speaker, and a model based on measurements to predict that. I guess if you wanted to design a speaker that had high "BassPunch," this paper would be a start on how to go about that. Good luck trying to market such a concept though.

Furthermore, such sub-spectrum analysis is also used in Harman listening tests just as well. This is how data on importance of bass performance was generated for example.

As to your personal note, I have listened and participated in Harman testing twice. I have first-hand feel for it. I imagine you have no such sense about this test. Nor have you read it from what I can tell. Just handing things to use to analyze instead of doing your own homework to understand the scope of a test and its relevance.

And oh, Genelec speakers were used as anchor in the study. May that be the reason it is on your radar???

I found the Volk, Bech, Pedersen, Christensen and Flemming paper using the search function on AES’s journal archive page.

I thought this comment was of interest:

«Loudspeaker specifications have traditionally described the physical properties and characteristics of loudspeakers: frequency response, dimensions and volume of the cabinet, diameter of drivers, impedance, total harmonic distortion, sensitivity, etc. Few of these directly describe the sound reproduction and none directly describe perception of the reproduction, i.e., takes into account that the human auditory system is highly non-linear in terms of spectral-, temporal-, and sound level processing (see, e.g., [3]). This disconnect between specifications and perception have made it challenging for acousticians and engineers (and consumers) to predict how a loudspeaker will sound on the basis of these specifications».

I am surprised you don’t share any of the authors’ observations that traditional descriptors in many cases do not «directly describe perception of the reproduction».

The authors note that none of the traditional descriptors «takes into account that the human auditory system is highly non-linear in terms of spectral-, temporal-, and sound level processing».

Much audio research bypasses a more fundamental model and understanding of human perception. The Harman model seems to be preference oriented, and not so much digging into models of human perception that you’d find in medical schools etc. Maybe the authors being more open to other schools of thought make them see the world differently than you do?
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
I am surprised you don’t share any of the authors’ observations that traditional descriptors in many cases do not «directly describe perception of the reproduction».

I don't remember he said that. In fact, I don't remember anybody said that in this topic. Can you post a quote when he said such thing?
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
I don't remember he said that. In fact, I don't remember anybody said that in this topic. Can you post a quote when he said such thing?

The article I quoted says:

«Few of these [traditional descriptors/factors] directly describe the sound...»

The authors also wrote that «[t]his disconnect between specifications and perception have made it challenging for acousticians and engineers (and consumers) to predict how a loudspeaker will sound on the basis of these specifications».

I don’t have the impression that @amirm shares those views.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
The article I quoted says:

«Few of these [traditional descriptors/factors] directly describe the sound...»

The authors also wrote that «[t]his disconnect between specifications and perception have made it challenging for acousticians and engineers (and consumers) to predict how a loudspeaker will sound on the basis of these specifications».

I don’t have the impression that @amirm shares those views.

Can you quote based on what you have such impression?

As I said, I don't remember not only Amir but anybody else said anything that would sound like opposing to that quote ("Few of these..").

Actually, I remember quite the opposite. Concept of Spinorama measurements (which practically all of us agree with) was designed to address that issue.
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom

I don't think I need to tell you that third-party independent anechoic measurements done in rigorous conditions from appropriate sources (such as the one I linked, which is from NRC) of course trump the manufacturer's published specifications. According to actual measurements, Devialet's claim of +/- 2 dB frequency response accuracy is nothing but wishful thinking.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
Can you quote based on what you have such impression?

As I said, I don't remember not only Amir but anybody else said anything that would sound like opposing to that quote ("Few of these..").

Actually, I remember quite the opposite. Concept of Spinorama measurements (which practically all of us agree with) was designed to address that issue.

Not long ago, @amirm wrote:

«Again, not at all. In the words of Sean Olive, the top 3 important parameters for any speaker are frequency response, frequency response and frequency response!»
Source: https://www.audiosciencereview.com/...he-last-twenty-years.8021/page-14#post-197155
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
Not long ago, @amirm wrote:

«Again, not at all. In the words of Sean Olive, the top 3 important parameters for any speaker are frequency response, frequency response and frequency response!»
Source: https://www.audiosciencereview.com/...he-last-twenty-years.8021/page-14#post-197155

That was said in the context of spinorama frequency response measurements, not a simple ± 2dB from 20Hz to 20kHz like in Devialet Phantom spec. And, unlike a simple on-axis frequency response, spinorama measurements have proven to show pretty good correlation with listener perception.

You, on the other hand, don't seem to agree with spinorama concept yet you have failed to show any arguments nor citations for a better solution.
 

Floyd Toole

Senior Member
Audio Luminary
Technical Expert
Industry Insider
Forum Donor
Joined
Mar 12, 2018
Messages
367
Likes
3,893
I remember the predictions from spin-o-rama results on bookshelf speakers without response below 100hz was close to 100% correlation. I don't know, maybe @Floyd Toole could answer about why. I'd assumed it was because at lower frequencies in their test rooms the response of the speaker was more corrupted by room modes and uneven response even when the speaker was fine in an anechoic chamber. That would fit with FR being even more of a direct indicator of how well quality of a loudspeaker is judged in listening tests.

I also wonder if there are thoughts on why something like the Quad's mentioned earlier seem to have a very high appeal to those who hear them while not being that good? Or maybe if it is anything more than the novel appearance, operation and expectation of the speaker without a box?

If you read the Olive AES papers or the summary in my book, it is explained that bass performance accounts for about 30% of the factor weighting in sound quality evaluations. There were two tests, one with 70 loudspeakers of all sizes, prices and bass extension (bandwidth). The correlation coefficient between double-blind subjective ratings and predicted ratings was 0.86. In a separate test using 13 bookshelf loudspeakers having similar bass extension, the correlation coefficient was 0.995. It was assumed that the higher correlation coefficient was at least partly due to the fact that the greatly varying bass extension of the 70 tested loudspeakers was not a factor. The fact that this much smaller test was done in a short time period, not the many months it took to collect the 70-speaker data, was also a probable factor.

The room, of course, is always a factor at low frequencies, but in these subjective evaluations were done in the Harman positional-substitution listening room - all active loudspeakers are brought to the same location in the same room and the listeners were always in the same seat. The room is there, but it was a constant factor. The positional exchange among the loudspeakers (tested four at a time) takes about 3 s, and this alone is a huge factor in achieving repeatable, trustworthy subjective data.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
That was said in the context of spinorama frequency response measurements, not a simple ± 2dB from 20Hz to 20kHz like in Devialet Phantom spec. And, unlike a simple on-axis frequency response, spinorama measurements have proven to show pretty good correlation with listener perception.

You, on the other hand, don't seem to agree with spinorama concept yet you have failed to show any arguments nor citations for a better solution.

My own speakers, Genelec 8351, are designed to be flat and smooth. In fact, their horizontal and vertical directivity measurements are probably the finest in the world.

The company that made my speakers, Genelec, are among the speaker builders in the world that follow the «Toole» book the closest. No wonder Genelec’s founder, Ilpo Martikainen recommended @Floyd Toole ‘s book warmly to interested customers. So I have personal experience with «correct» speakers. I know the logic behind such designs.

However, things become interesting when you compare speakers side by side, like 8351 vs 8341 vs 8331 (41 and 31 are smaller designs of 8351), Kii Three and Genelec S360. On paper, these speakers look similar. But do they sound the same?

I am the first person usually to point out that speakers are «boring» in the sense that I think good speakers have more similarities than obvious differences. But it doesn’t mean I think 8331, 8341, 8351, S360 and Kii Three sound exactly the same even if their specifications and measurements are similar. Even if this forum wants me to shut up, fall into the fold and repeat the gospel that correct speakers all sound the same, I will not. Should I be sorry for perceiving differences between 8331, 8341, 8351, S360 and Kii Three (i.e. speakers that I have heard in the same room)?

The flat and smooth factor is an obvious parameter around which competent speaker designers make speakers. My point and intent was to think beyond the «flat and smooth» factor, to investigate if there’s more. I referred to a JAES paper that asked why there’s a disconnect between specifications and perception. Simply asking this question makes some people at ASR herd like a cult when they sense a dissident. And that’s why I also stand by the impression that audio science lacks the maturity seen in other sciences. Where’s the curiosity? Where’s the interest for cross-disciplinar science and research? And importantly: Where’s the humility when faced with a complex issue?
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,374
Likes
234,446
Location
Seattle Area
I referred to a JAES paper that asked why there’s a disconnect between specifications and perception. Simply asking this question makes some people at ASR herd like a cult when they sense a dissident.
Nothing like that happened. The reaction was your malposition of the paper against Harman research when the two were not addressing the same thing. That work is not a counter to Harman's research as you implied.

But yes, to the extent your posts come across as protests, you are going to get emotion involved in the responses you get.

I referred to a JAES paper that asked why there’s a disconnect between specifications and perception.
See? That is not what the paper is about. Read the thing before summarizing such. The abstract of a paper is not enough. As I mentioned, they want to create new vocabulary to describe tonality of speakers. It has nothing to do with "perception and specifications."
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,374
Likes
234,446
Location
Seattle Area
I thought this comment was of interest:

«Loudspeaker specifications have traditionally described the physical properties and characteristics of loudspeakers: frequency response, dimensions and volume of the cabinet, diameter of drivers, impedance, total harmonic distortion, sensitivity, etc. Few of these directly describe the sound reproduction and none directly describe perception of the reproduction, i.e., takes into account that the human auditory system is highly non-linear in terms of spectral-, temporal-, and sound level processing (see, e.g., [3]). This disconnect between specifications and perception have made it challenging for acousticians and engineers (and consumers) to predict how a loudspeaker will sound on the basis of these specifications».
This "disconnect" has been addressed extensively through the work of Dr. Toole, Olive, etc.

What they are talking about is what we all agree with: that the marketing specs for a speaker doesn't tell you how it sounds. Well duh! We know that. This is why much more data needs to be provided with respect to on and off-axis of speaker. This data as proposed by researchers and standardized in CEA 2034 requires high resolution frequency response measurements (at 2 Hz resolution) and 70 or so measurements all around the speakers. Once there, we can make very strong predictions. No one is talking about empty frequency response specs from speaker manufacturers.

In addition to graphs, Harman research almost routinely includes subjective comments from listeners, categorization of response in different bands (e.g. bass), etc.
 
OP
N

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,110
Likes
8,433
Location
NYC
My own speakers, Genelec 8351, are designed to be flat and smooth. In fact, their horizontal and vertical directivity measurements are probably the finest in the world.

The company that made my speakers, Genelec, are among the speaker builders in the world that follow the «Toole» book the closest. No wonder Genelec’s founder, Ilpo Martikainen recommended @Floyd Toole ‘s book warmly to interested customers. So I have personal experience with «correct» speakers. I know the logic behind such designs.

However, things become interesting when you compare speakers side by side, like 8351 vs 8341 vs 8331 (41 and 31 are smaller designs of 8351), Kii Three and Genelec S360. On paper, these speakers look similar. But do they sound the same?

I am the first person usually to point out that speakers are «boring» in the sense that I think good speakers have more similarities than obvious differences. But it doesn’t mean I think 8331, 8341, 8351, S360 and Kii Three sound exactly the same even if their specifications and measurements are similar. Even if this forum wants me to shut up, fall into the fold and repeat the gospel that correct speakers all sound the same, I will not. Should I be sorry for perceiving differences between 8331, 8341, 8351, S360 and Kii Three (i.e. speakers that I have heard in the same room)?

The flat and smooth factor is an obvious parameter around which competent speaker designers make speakers. My point and intent was to think beyond the «flat and smooth» factor, to investigate if there’s more. I referred to a JAES paper that asked why there’s a disconnect between specifications and perception. Simply asking this question makes some people at ASR herd like a cult when they sense a dissident. And that’s why I also stand by the impression that audio science lacks the maturity seen in other sciences. Where’s the curiosity? Where’s the interest for cross-disciplinar science and research? And importantly: Where’s the humility when faced with a complex issue?

Sorry, but I don't understand. When has anyone said that all good speakers must sound the same?

I do not think anyone here expects or wants all speakers to sound 100 percent identical. We're simply highlighting the fact that research points to frequency response being the number one factor in listener preference, and that it's curious some manufacturers seem to actively go against this science.

I started this thread roughly asking: "if research shows flat on axis and smooth directivity is strongly correlated to listener preference, why do some manufacturers actively design their speakers otherwise?"

We know that starting with good frequency response on and off axis is good. There is not research I'm aware that suggests otherwise - and manufacturer specs are of limited use unless they include comprehensive anechoic data.

If a manufacturer diverts from good frequency responsefor reasons beyond budget, it better have a really good reason why. But that doesn't mean speakers that do have good frequency response can't try to optimize other parameters to give themselves an edge. It just seems they generally should be prioritizing frequency response before worrying about other things.
 
Last edited:

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
Does the ideal of "straight wire with gain" apply to the whole system?
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
Nothing like that happened. The reaction was your malposition of the paper against Harman research when the two were not addressing the same thing. That work is not a counter to Harman's research as you implied.

But yes, to the extent your posts come across as protests, you are going to get emotion involved in the responses you get.


See? That is not what the paper is about. Read the thing before summarizing such. The abstract of a paper is not enough. As I mentioned, they want to create new vocabulary to describe tonality of speakers. It has nothing to do with "perception and specifications."

@amirm , you wrote:

«See? That is not what the paper is about. Read the thing before summarizing such. The abstract of a paper is not enough. As I mentioned, they want to create new vocabulary to describe tonality of speakers. It has nothing to do with "perception and specifications."»

I don’t quite get what you’re writing here. Let me show you what is the intention of the authors using an article to can be read as a run-up to the JAES article (https://assets.madebydelta.com/assets/docs/senselab/publications/TEKnotat_TN11_UK_v5.pdf).

Take a look at the excerpt below:

B5A9E405-3729-40A7-B04F-300BDA28FB17.png

Source: https://assets.madebydelta.com/assets/docs/senselab/publications/TEKnotat_TN11_UK_v5.pdf

For good order I will paste the text above here (my underlining):

«Interesting relationships
Figure 18 shows that in general, there is no relationship between the experienced bass depth and the technical data for the low frequency limit. Four of the loudspeakers more or less had the same lower frequency limit of 50–53 Hz, yet they were assessed as having a large (and significant) difference in the perceived Bass Strength. The same applies to DALI Menuet and DALI Opticon 2, whose respec- tive bass depth was assessed as being quite different, while the data stated that both had a lower frequency limit of 59 Hz. The fact that you cannot assess a loudspeaker’s sound based on the tech- nical data is hardly new, however it is still interesting to see it demonstrated with a well-defined listen- ing test. In other words, if you want to know how a loudspeaker sounds, it is more sensible to use a perceptual assessment of a loudspeaker’s sound based on a listening test rather than taking outset in the technical data.»

The author(s) makes the point that speaker sound could follow the example of wine and beer reviews:

DD874A58-334A-49F9-A583-E084E0C4A422.jpeg


So specifications and perception are obviously what interests these authors.

FWIW, the authors also look at speaker volume and bass quality, as well as preference vs price:

B3E875CF-52C9-4EC2-8464-7C395A1E0819.png


You may dislike what the authors write, but I cannot see I misrepresented the underlying intention and opinion of the authors.
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
Even if this forum wants me to shut up, fall into the fold and repeat the gospel that correct speakers all sound the same, I will not. Should I be sorry for perceiving differences between 8331, 8341, 8351, S360 and Kii Three (i.e. speakers that I have heard in the same room)?

Toole & Olive's research does not predict all speakers with "correct" responses sound the same. It only predicts they will get similar preference ratings (that is, good ratings). They might sound different, but they are unlikely to sound different in ways that would make their preference ratings diverge.

Also, keep in mind that this research inherently deals with statistics and averages. It attempts to predict listener preference for the average listener in the average room. That's another reason why you don't have to feel "sorry" for perceiving differences.

And even if we cast that aside, keep in mind the correlation is not perfect. Olive's listener preference prediction model has an accuracy of 0.91. That's close to 1, but it's not 1. That still leaves margin for error, especially when the loudspeakers are already very close to each other (such as with the Genelecs you mention).

You seem to be attacking some kind of strawman who claims that Toole/Olive's research is perfect and will perfectly predict how any loudspeaker will be perceived by any listener in any room. Of course that's not the case. But it's still the best tool we have to make informed purchase recommendations that are most relevant to a majority of people in a majority of rooms.
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,253
Toole & Olive's research does not predict all speakers with "correct" responses sound the same. It only predicts they will get similar preference ratings (that is, good ratings). They might sound different, but they are unlikely to sound different in ways that would make their preference ratings diverge.

Also, keep in mind that this research inherently deals with statistics and averages. It attempts to predict listener preference for the average listener in the average room. That's another reason why you don't have to feel "sorry" for perceiving differences.

And even if we cast that aside, keep in mind the correlation is not perfect. Olive's listener preference prediction model has an accuracy of 0.91. That's close to 1, but it's not 1. That still leaves margin for error, especially when the loudspeakers are already very close to each other (such as with the Genelecs you mention).

You seem to be attacking some kind of strawman who claims that Toole/Olive's research is perfect and will perfectly predict how any loudspeaker will be perceived by any listener in any room. Of course that's not the case. But it's still the best tool we have to make informed purchase recommendations that are most relevant to a majority of people in a majority of rooms.

My «strawman» is the marketeers of certain «correct» mini speakers who cite specifications that defy the laws of physics. So my «strawman» is real. I have shown examples.

See also my post #297 above for an alternative speaker evalutation method that is an alternative to the purely measurement driven method. This kind of research suggests real speaker reviews should be a hybrid of objective measurements and subjective blind tests.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,374
Likes
234,446
Location
Seattle Area
You may dislike what the authors write, but I cannot see I misrepresented the underlying intention and opinion of the authors.
I have no dislike for that article which I have already read. As I explained about the AES version, it is a decent effort to create a vocabulary for sound. That it criticizes simple frequency response curves is not new or controversial. You actually quoted the authors saying the same: "The fact that you cannot assess a loudspeaker’s sound based on the tech- nical data is hardly new, however it is still interesting to see it demonstrated with a well-defined listen- ing test. "

In no way does it critique the work we are discussing with full spinorama and correlation of that with countless listening tests. Similar observations are documented in detail by work of Olive, etc. You just haven't read them or understand them so continue to think this is somehow new. It is different but not new.

Instead of googling, spend a few dollars and buy the actual AES papers we and you are discussing. If you don't understand them, ask questions. Otherwise you continue to waste our time with these protests.
 
Top Bottom