• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

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

MRC01

Senior Member
Joined
Feb 5, 2019
Messages
389
Likes
259
#61
I think panel speakers exaggerate treble to make up for the very narrow directivity in this region, just to get some HF into the room.
Panel speakers really damage the reputation of dipoles. Dipole dynamic speakers can do things no other speakers can. RIP Siegfried.
If by panel you mean electrostat, that can happen when the wavelength of treble frequencies is shorter than the physical width of the driver, which causes an uneven beamy dispersion. However, if the panel speaker has a ribbon tweeter (like Magnepan), the treble dispersion is more uniform. And the ribbon tweeters have low distortion (mine measure around -60 dB in-room) with reasonably smooth/flat response.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#62
Technically correct speakers make good references for evaluating recordings but they are not necessarily the most musically enjoyable speakers to everyone. Yes, average listeners in laboratory settings prefer a flat and smooth frequency response, but many audiophiles see the sound produced by their system as something which they have built to explore music according to their taste and understanding. Musical taste has a lot to do with it - how can one 'correct' speaker be ideal for everything if recordings vary so much? In this sense, a technically perfect speaker becomes a jack of all trades and a master of few. If you listen to ancient jazz and blues recordings, or solely to classical, or modern electronic music, the ideal speaker will vary. A speaker with low distortion, flat FR and smooth directivity transitions will probably not sound terrible on many recordings, but we can't really blame people for liking the sound of dipoles, or harbeths, or horns, altec coaxials, line arrays, and any other number of fundamentally compromised but beloved speakers.

It is not simply an issue of marketing or the incompetence of the designers. Audiophiles simply have different goals than neutrality, and speaker designers provide them with options.
Two things here. The Toole et al research shows that people's tastes don't vary as much as you may think. That's precisely why the technical concensus of flat on axis and smooth off axis response is reached. The blind subjective listening tests with lots of different people tell us so.

However, yes you are absolutely correct in saying that recordings vary enormously in tonality etc. IIRC Toole advocates the use of variable tone controls to ameliorate this variability.

Fixing the sound with coloured speaker doesn't make sense. It will only sound correct (good) with a limited range of recording sound.

FWIW I have found that the more neutral the speaker is, the less I think about the sound of the recording and the more I get involved with the music. Also I find fewer recordings have objectionable sound quality.
 
Last edited:

MRC01

Senior Member
Joined
Feb 5, 2019
Messages
389
Likes
259
#63
... I have found that the more neutral the speaker is the less I think about the sound of the recording and the more I get involved with the music. Also I find fewer recordings have objectionable sound quality.
I find that better (more accurate, more neutral) playback systems, enhance the differences between recordings. That is, different recordings sound more different. Yet ironically, a really good playback system doesn't make imperfect or flawed recordings un-listenable. The flaws are clearly evident, but they usually aren't ear-jarring.

It's not uncommon to hear audiophiles describe high resolution systems as sounding great with a select handful of recordings, and making other recordings sound terrible or truly ear-jarring. In these cases, I suspect the system, not the recordings. That is, it may be the case that the allegedly "high resolution" system has specific flaws that are being alternately mitigated or exaggerated by different recordings, where a more neutral truly high resolution system would render these various recordings in a more fair light.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
6,352
Likes
5,941
#64
I find that better (more accurate, more neutral) playback systems, enhance the differences between recordings. That is, different recordings sound more different. Yet ironically, a really good playback system doesn't make imperfect or flawed recordings un-listenable. The flaws are clearly evident, but they usually aren't ear-jarring.

It's not uncommon to hear audiophiles describe high resolution systems as sounding great with a select handful of recordings, and making other recordings sound terrible or truly ear-jarring. In these cases, I suspect the system, not the recordings. That is, it may be the case that the allegedly "high resolution" system has specific flaws that are being alternately mitigated or exaggerated by different recordings, where a more neutral truly high resolution system would render these various recordings in a more fair light.
Some of these high resolution systems have incredibly flawed speakers. It all comes back to letting your vision corrupt your hearing along with everything else. I stole this from someone else. The bandwidth of sighted listening.

1563316728689.png


Sighted listening has a bandwidth of 340 terahertz. Blind listening only uses 20,000 hertz.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#65
I find that better (more accurate, more neutral) playback systems, enhance the differences between recordings. That is, different recordings sound more different. Yet ironically, a really good playback system doesn't make imperfect or flawed recordings un-listenable. The flaws are clearly evident, but they usually aren't ear-jarring.

It's not uncommon to hear audiophiles describe high resolution systems as sounding great with a select handful of recordings, and making other recordings sound terrible or truly ear-jarring. In these cases, I suspect the system, not the recordings. That is, it may be the case that the allegedly "high resolution" system has specific flaws that are being alternately mitigated or exaggerated by different recordings, where a more neutral truly high resolution system would render these various recordings in a more fair light.
Broadly my experience too.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#66
Two things here. The Toole et al research shows that people's tastes don't vary as much as you may think. That's precisely why the technical concensus of flat on axis and smooth off axis response is reached. The blind tests with lots of different people tell us so.

However, yes you are absolutely correct in saying that recordings vary enormously in tonality etc. IIRC Toole advocates the use of variable tone controls to ameliorate this variability.

Fixing the sound with coloured speaker doesn't make sense. It will only sound correct (good) with a limited range of recording sound.

FWIW I have found that the more neutral the speaker is, the less I think about the sound of the recording and the more I get involved with the music. Also I find fewer recordings have objectionable sound quality.
The other issue to note is that until such time that there are proper standards for recording studio monitoring we are going to continue to be plagued by wonky sounding recordings.
 

oivavoi

Major Contributor
Joined
Jan 12, 2017
Messages
1,230
Likes
663
#67
In order to assess the impact of controlled directivity, it would be interesting to have a repeat of the Harman blind test but including speakers design (or should i say marketed) with different chief design approaches and at least share very good on-axis response, for example:

Controlled directivity as chief design trait
- Revel Salon 2
- KEF reference 5 (uni drivers)

Non controlled directivity as chief design trait
- ATC scm 100 (low distortion)
- Dunlavy (phase accuracy, perfect step response).
- B&W (?)
I agree with this. The most important thing, I think, is actually to get other research groups to replicate the experiments, and see if they come to the same results. I do research for a living, albeit in a field which has nothing to do with audio at all. And in the fields in which I work or which I follow closely, you can't take findings for granted unless they are replicated and corroborated by other research groups. There is a reason people speak about a "replication crisis" in science... scientists often get small things wrong, even though they act in good faith, and it's difficult to know whether one has actually measured what one thinks one has measured. One of the most important things is actually to have different research groups working on similar questions, as this makes it more likely that mistakes or limitations are discovered.

What troubles me slightly with lifting up the Harman research to the status of an undisputed audio gospel is that this is basically one research group, who also had commercial interests in what they were/are doing. I don't mean to disparage their work at all - Toole, Olive and Welti and the others have done an immense service to the audio community by putting so much of their work out in the public domain. Really. The problem is not that they have been doing this work - quite on the contrary - it's rather that so few others have bothered to replicate or challenge the studies! "Psychoacoustic loudspeaker science" is basically such a small and underfunded field... and unfortunately it's probably going to stay that way, given that high-end audio seems to be destined for the same end as the dinosaurs.

But take the issue of the dipole speaker, for example. In Harman studies, no good. In one of Søren Bech's studies, however, the dipole received the highest rating given a particular placement, and the worst rating, given a different placement.
bech directivity.JPG
bech directivity.JPG

Source: https://www.researchgate.net/public..._sound_quality_-_a_review_of_existing_studies

This is just an example. Now it would be relatively stupid to claim that flat frequency response on-axis and smooth response off-axis is bad. Nobody thinks that. The interesting devil lies in all the details in speaker design. How much weight should on-axis flatness be given, relative to behavior off-axis? There is a genuine trade-off here - equalizing the speaker to improve off-axis may make it slightly worse on-axis. Waveguides or horns may improve directivity patterns, but how about those HOMs? And what about things like transient response, dynamics etc? (here's a paper I read some time ago on that: https://eprints.soton.ac.uk/426110/ )

I don't mean to spread FUD here. The Harman way of doing speakers is obviously a very good and valid way of doing speakers. In my sighted listening I usually tend to like speakers that are designed according to the Harman philosophy. But my sighted listening has also told me that I can like equally well speakers that are designed according to other philosophies (prioritizing low distorsion, phase coherencey and "flat enough" response on-axis, for example). Until other research groups start replicating Harman's experiments, I'll probably not become convinced that flat response on-axis and smooth response off-axis are the only things that matter.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
20,088
Likes
23,264
Location
Seattle Area
#69
What troubles me slightly with lifting up the Harman research to the status of an undisputed audio gospel is that this is basically one research group, who had clear commercial interests in what they were/are doing.
Vast amount of this research came under the umbrella of National Research Council of Canada. That is where Dr. Toole, Olive, etc. did their research and published countless papers on the topic. Harman allowed expansion of that work a decade or two later but the conclusions of what we are talking about are all sourced in the NRC work. So the commercial interest angle doesn't hold.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#70
I agree with this. The most important thing, I think, is actually to get other research groups to replicate the experiments, and see if they come to the same results. I do research for a living, albeit in a field which has nothing to do with audio at all. And in the fields in which I work or which I follow closely, you can't take findings for granted unless they are replicated and corroborated by other research groups. There is a reason people speak about a "replication crisis" in science... scientists often get small things wrong, even though they act in good faith, and it's difficult to know whether one has actually measured what one thinks one has measured. One of the most important things is actually to have different research groups working on similar questions, as this makes it more likely that mistakes or limitations are discovered.

What troubles me slightly with lifting up the Harman research to the status of an undisputed audio gospel is that this is basically one research group, who had clear commercial interests in what they were/are doing. I don't mean to disparage their work at all - Toole, Olive and Welti and the others have done an immense service to the audio community by putting so much of their work out in the public domain. Really. The problem is not that they have been doing this this work - quite on the contrary - it's rather that so few others have bothered to replicate or challenge the studies! "Psychoacoustic loudspeaker science" is basically such a small and underfunded field... and unfortunately it's probably going to stay that way, given that high-end audio seems to be destined for the same end as the dinosaurs.

But take the issue of the dipole speaker, for example. In Harman studies, no good. In one of Søren Bech's studies, however, the dipole received the highest rating given a particular placement, and the worst rating, given a different placement.
View attachment 29560 View attachment 29560
Source: https://www.researchgate.net/public..._sound_quality_-_a_review_of_existing_studies

This is just an example. Now it would be relatively stupid to claim that flat frequency response on-axis and smooth response off-axis is bad. Nobody thinks that. The interesting devil lies in all the details in speaker design. How much weight should on-axis flatness be given, relative to behavior off-axis? There is a genuine trade-off here - equalizing the speaker to improve off-axis may make it slightly worse on-axis. Waveguides or horns may improve directivity patterns, but how about those HOMs? And what about things like transient response, dynamics etc? (here's a paper I read some time ago on that: https://eprints.soton.ac.uk/426110/ )

I don't mean to spread FUD here. The Harman way of doing speakers is obviously a very good and valid way of doing speakers. In my sighted listening I usually tend to like speakers that are designed according to the Harman philosophy. But until other research groups start replicating Harman's experiments, I'll probably not become convinced that flat response on-axis and smooth response off-axis are the only things that matter.
I certainly take your point regarding the replication of research, however we know that other manufacturers do work to the same rationale because their own research concurs.

See the example of KEF above.

My question would be in that case, what are the conflicting philosophies for speaker design? What is a different way of doing things? Why is it better and what alternative research supports it?
 

MRC01

Senior Member
Joined
Feb 5, 2019
Messages
389
Likes
259
#71
... take the issue of the dipole speaker, for example. In Harman studies, no good. In one of Søren Bech's studies, however, the dipole received the highest rating given a particular placement, and the worst rating, given a different placement.
...
I'm not surprised. IME, dipoles are more sensitive than conventional speakers, to room setup and location. Pragmatically, this is both a blessing and a curse.
 

oivavoi

Major Contributor
Joined
Jan 12, 2017
Messages
1,230
Likes
663
#72
Vast amount of this research came under the umbrella of National Research Council of Canada. That is where Dr. Toole, Olive, etc. did their research and published countless papers on the topic. Harman allowed expansion of that work a decade or two later but the conclusions of what we are talking about are all sourced in the NRC work. So the commercial interest angle doesn't hold.
That's true. But it would nevertheless be a bit strange if commercial interests played no role at all for the later research, considering what questions they thought it was worth looking into. If it played absolutely no role, then the Harman researchers would be the first researchers in the history of science who would not be affected by their funding at all... (I say this as a person who has done a lot of commissioned research, and who knows how easy it is to become affected by such things unconsciously).

An example: Given that the quasi-omni mirage speaker was the highest-ranking speaker at the NRC ever, how come Harman didn't look into that design-type more in the spinoramas? I would guess that it's because such omni speakers require a placement which is not very living room friendly. If they did compare their own Revel speakers with Mirage-like designs, I'll need to eat my hat though! :)
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#73
I'm not surprised. IME, dipoles are more sensitive than conventional speakers, to room setup and location. Pragmatically, this is both a blessing and a curse.
It's an interesting point. When does a dipole work well?
I will be the first to admit that I have next to no experience of dipoles. Being devil's advocate, do they work well when the negative effects of being a dipole are mitigated by the room and position.? What I mean by that is when they are prevented from behaving too much like dipoles and more like conventional box speakers ;)
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#74
That's true. But it would nevertheless be a bit strange if commercial interests played no role at all for the later research, considering what questions they thought it was worth looking into. If it played absolutely no role, then the Harman researchers would be the first researchers in the history of science who would not be affected by their funding at all... (I say this as a person who has done a lot of commissioned research, and who knows how easy it is to become affected by such things unconsciously).

An example: Given that the quasi-omni mirage speaker was the highest-ranking speaker at the NRC ever, how come Harman didn't look into that design-type more in the spinoramas? I would guess that it's because such omni speakers require a placement which is not very living room friendly. If they did compare their own Revel speakers with Mirage-like designs, I'll need to eat my hat though! :)
But what is the commercial interest? It is to make the best sounding speaker. So what is the conflict?

Obviously there is more to commercial marketing than just the sound quality. We all know audiophiles buy with their eyes and are influenced by other things, but what would be the commercial sense in making a speaker that wasn't liked for its audio quality?

IMO the technical and commercial objectives are not contradictory.
 

oivavoi

Major Contributor
Joined
Jan 12, 2017
Messages
1,230
Likes
663
#75
I certainly take your point regarding the replication of research, however we know that other manufacturers do work to the same rationale because their own research concurs.

See the example of KEF above.

My question would be in that case, what are the conflicting philosophies for speaker design? What is a different way of doing things? Why is it better and what alternative research supports it?
Here's what I think are real dilemmas, where the Harman/Toole research may have left matters unresolved:

- Wide vs narrow directivity. Toole explicitly says wide directivity is good. Geddes thinks narrow directivity is better, and apparently has a paper coming out about it.

- Off-axis vs on-axis: Equalizing for better off-axis response may make the speakers less flat on-axis. How much weight should one give to one vs the other? Toole/Harman made a formula for this (listening window etc), but I'm not sure it's necessary valid for all kinds of listening situations.

- Point source vs many-ways: Here there's a trade-off between the point source ideal and things like distorsion, dynamics etc.

- Limits of eq: Does it matter how flat the speakers/drivers are before equalization? Or can everything be equalized digitally to the same frequency response, and it will sound the same? An example: Speaker A plays perfectly out of the box, without eq. Speaker B plays horribly out of the box, but digital EQ makes it measure like speaker A. Does this matter for the end result?

Just some brief thoughts, it's getting late in Norway...
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#76
Here's what I think are real dilemmas, where the Harman/Toole research may have left matters unresolved:
Wide vs narrow directivity. Toole explicitly says wide directivity is good. Geddes thinks narrow directivity is better, and apparently has a paper coming out about it.

Off-axis vs on-axis: Equalizing for better off-axis response may make the speakers less flat on-axis. How much weight should one give to one vs the other? Toole/Harman made a formula for this (listening window etc), but I'm not sure it's necessary valid for all kinds of listening situations.

Point source vs many-ways: Here there's a trade-off between the point source ideal and things like distorsion, dynamics etc.

Limits of eq: Does it matter how flat the speakers/drivers are before equalization? Or can everything be equalized digitally to the same frequency response, and it will sound the same? An example: Speaker A plays perfectly out of the box, without eq. Speaker B plays horribly out of the box, but digital EQ makes it measure like speaker A. Does this matter for the end result?
Well we need to see the Geddes research and how it stacks up. Until then.....

OK, why do you think that keeping on axis flat is the wrong thing to do?

EQ, if it a minimum phase issue then it can be dealt with. If its a directivity problem then not.

I think most of your questions here are actually answered pretty clearly in the @Floyd Toole publications, but hopefully he might be able to comment directly.
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
20,088
Likes
23,264
Location
Seattle Area
#77
That's true. But it would nevertheless be a bit strange if commercial interests played no role at all for the later research, considering what questions they thought it was worth looking into.
The commercial interest was to then design better speakers and car audio systems to win more business in the future. It was not as you imply to defend existing designs. Indeed a battle ensued when Dr. Toole and Olive joined harman, claiming their research was wrong. And that they should make different speakers for German audiences than others. A controlled listening test was created which showed this to be clearly wrong.

Some other groups inside Harman such as JBL continued to work against this research until the last few years.

The common example of this is companies hiring top researchers out of university to work for them to develop future products. Doing so doesn't at all show biased influence to create bad products.
 

oivavoi

Major Contributor
Joined
Jan 12, 2017
Messages
1,230
Likes
663
#78
Well we need to see the Geddes research and how it stacks up. Until then.....

OK, why do you think that keeping on axis flat is the wrong thing to do?

EQ, if it a minimum phase issue then it can be dealt with. If its a directivity problem then not.

I think your questions here are actually answered pretty clearly in the @Floyd Toole publications, but hopefully he might be able to comment directly.
Clarification: No, I don't think flat on-axis is wrong at all. What I wonder about is the trade-off when equalizing for off-axis smoothness or flatness, which may make it slightly less flat on-axis. My hunch is that when listning close to the speakers far away from walls for example, it would be better to keep on-axis as flat as possible, even with slightly worse off-axis response.

I have read most of dr. Toole's publications, and find them indispensable for me as an audiophile. As I said, his work, both at NRC and Harman, was an immense service to the audio community. But I actually didn't tag him here on purpose, because my point was not to "challenge" him. I'm no technical expert at all! It was more a general point that it's prudent to keep some level of open-mindedness in a field where the replications are still so few.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#79
Clarification: No, I don't think flat on-axis is wrong at all. What I wonder about is the trade-off when equalizing for off-axis smoothness or flatness, which may make it slightly less flat on-axis. My hunch is that when listning close to the speakers far away from walls for example, it would be better to keep on-axis as flat as possible, even with slightly worse off-axis response.

I have read most of dr. Toole's publications, and find them indispensable for me as an audiophile. As I said, his work, both at NRC and Harman, was an immense service to the audio community. But I actually didn't tag him here on purpose, because my point was not to "challenge" him. I'm no technical expert at all! It was more a general point that it's prudent to keep some level of open-mindedness in a field where the replications are still so few.
It's not to challenge him, it's just to provide further clarification and insight into the work and conclusions. It's fantastic to have his contribution.
 

oivavoi

Major Contributor
Joined
Jan 12, 2017
Messages
1,230
Likes
663
#80
The commercial interest was to then design better speakers and car audio systems to win more business in the future. It was not as you imply to defend existing designs. Indeed a battle ensued when Dr. Toole and Olive joined harman, claiming their research was wrong. And that they should make different speakers for German audiences than others. A controlled listening test was created which showed this to be clearly wrong.

Some other groups inside Harman such as JBL continued to work against this research until the last few years.

The common example of this is companies hiring top researchers out of university to work for them to develop future products. Doing so doesn't at all show biased influence to create bad products.
I think you misunderstand me here Amir... I certainly didn't mean to imply that dr Toole and Olive defended existing designs. My point was rather that when the job was to create designs for the future, they inevitably had to make some choices about what could be commercially viable. Let's assume, hypothetically, that there were apriori reasons to think that a speaker that was four meters tall would be superior. Would they seriously consider that? I think not. I wouldn't have done so in their place, at least.
 
Top Bottom