What role should listening play in speaker design

[SigbergAudio]

Nobody said it is 100% random. You can't be twisting the argument so maliciously by mistake, or simply because you fell for Matt's debating tactics. In fact it is 100% non-random, in that one's cognitive biases will always work in the same directions at any one point in time (they will often evolve over time). There will be a lot of variations between individuals, but then again, there are often common threads among cognitive biases that result in many people being influenced in the same way by the same non-sonic variables. That's not '100% random'.

The (yes deliberately) exaggerated "100% random" claim, is to support the "sighted listening is 100% useless/pointless" argument. If this is the case, it is not very important if the effect is random or not, the result is the same.

Again, it would be helpful if we could stay with the context. The discussion is not the general / wide "can you trust your ears when auditioning speakers at a hifi show" and/or "are subjective loudspeaker reviews useless", but rather "is sighted listening during loudspeaker development pointless". So we are not comparing a fancy loudspeaker from a famous person against a boring looking speaker from an unknown Chinese company, we are comparing the same loudspeaker in different iterations, and the person doing the comparison is the person who designed it.

 

[Vuki]

It's not about what non linear response sounds best, but what linear response sounds best. When designing a speaker most of the time there is more then one way to achieve (almost) linear response (it is never absolutely linear) and choosing the way (compromise) is where listening is very useful.

 

[SigbergAudio]

How personal is your process ? You seems to be a one man operation for the moment .

Is it about what you prefer or what your customers prefer . Or what in some larger context is correct ( some posters here seems alluding to that this exist , but never defined it . Some say vaguely Tool and Olive some has an undisclosed personal standard like @Arindal ).

If you managed to crow your company ( I whish you well products seems amazing btw ) you better get that listening panel .

As mentioned earlier, other people are brought in on a regular basis to chime on what they hear, and even on how the measurements look.

With regards to what I prefer vs the customers prefer, I subscribe to the belief that our preferences are largely similar. There may be preferences (based on listening environment (room), experience, and/or genre preference that may result in deviation in preferred bass level. Fortunately I build active speakers, so I don't have to make a choice set in stone here, this can be easily adjusted by the customer.

 

[TimVG]

but rather "is sighted listening during loudspeaker development pointless

A fun experiment at some point would be to have someone design a crossover for one of your models based on nothing but the measurements and do a double-blind preference test.

 

[amirm]

It seems to me that those advocating speaker design w/o listening haven't designed any speaker ever. I could be wrong?

You are wrong. No one is saying don't listen. There are plenty of reasons to listen to a speaker. As I mentioned, even Harman uses blind listening tests with all the knowledge they have of the science.

What we are saying is that you better not use ad-hoc, sighted listening tests to override what you see in measurements. If measurements show flaws, use proper engineering to fix them.

All this aside, as someone who has compared measurements to sound of speakers across hundreds of them, I would say the reverse to you. If you are still an old school designer, optimizing the sound of a speaker with your ears per above, you will face a headwind when it comes to selling to us bunch. You can listen all you want. But don't screw up the measurements.

 

[SigbergAudio]

A fun experiment at some point would be to have someone design a crossover for one of your models based on nothing but the measurements and do a double-blind preference test.

Semi-related anecdote; Spinorama.org provides EQ suggestions to improve(?) all speakers that are there. The improvements made favors on-axis response over off-axis response. These are adjustments I could have made, but would for instance result in excess build-up in energy off-axis at some frequencies. These are adjustments I've obviously tested along the way, but have discarded due to me making other choices, and finding them to sound subjectively better.

I could be wrong and biased of course, hence my reference to this as an anecdote.

 

[amirm]

A fun experiment at some point would be to have someone design a crossover for one of your models based on nothing but the measurements and do a double-blind preference test.

That wouldn't tell us much as the issues may go beyond the crossover (think lack of waveguide). Or the crossover could become more expensive.

A more relevant comparison would be two speakers at the same price point, one using measurements as the metric of performance, and the other, tuning by ear, and showing worse measurements.

 

[Audionaut]

I mean my decent:
Not punishing impedance, linear, high SPL ability, low THD, etc.
I wouldn't even know where to begin on each of these attributes.
There's thousands of speaker brands, and a lot of them are nice, using the rights tools, etc.
How many nice driver companies are around? In the tens? A hundred? A tenth of them proven through time?
And how many of them build nice coaxials?

I believe that many engineers at major speaker companies could design a driver from scratch—based on their experience—that would come very close to the ideal.
There are a great many companies, especially in the PA sector, that manufacture very good coaxial drivers. If you’d like, I’d be happy to name a few.

All other times, on-axis correction ruled.
On-axis neutrality is the most important characteristic of a speaker.

I’d like to ask a couple of questions that seem to have been overlooked in all the back-and-forth here:

Does a loudspeaker with a linear frequency response curve sound the same at every volume level?

Does it sound the same at every listening distance?

Can anyone tell the difference between speakers with a linear on-axis frequency response, one of which uses a horn with a 60 x 40-degree dispersion pattern, versus an omnidirectional speaker, versus a coaxial speaker, versus a D’Appolito design?

Why are people arguing about whether a Neumann or Genelec monitor sounds “better”?

Can you hear a difference between a speaker with a linear on-axis frequency response curve that uses an 18-inch driver for bass, a 10-inch driver for midrange, and a compression driver in a horn for treble, versus a speaker that uses a 6-inch coaxial driver for mid-high frequencies and an 8-inch driver for bass?

Why are there so many different approaches to loudspeaker design—leaving visual aesthetics aside—when it’s now so easy to achieve a flat on-axis frequency response using DSP and solid amplifier technology?

Do these loudspeakers with a linear on-axis frequency response sound the same in every room?

Why are loudspeaker manufacturers still searching for the ideal diaphragm material today?


Personally, I believe that there are only approximations of the ideal loudspeaker, and that a loudspeaker must ALWAYS be suited to the room in which it is placed.

I can now describe my requirements (which are partly interdependent and overlap) for a largely idealized loudspeaker as follows:

Low distortion (harmonic distortion + intermodulation)

Volume stability = very high and undistorted volume possible when needed, without compression, without distortion

Consistent dynamic range across all volume levels = the speaker has excellent dynamic capabilities and reproduces very quiet passages just as cleanly as very high volumes. The soundstage remains consistent across all volume levels.

Very consistent directional characteristics (constant directivity)

Flat frequency response (on- and off-axis, smooth energy balance)

Good transient response = no time lag between drivers, no phase shift, clean impulse response of individual drivers, controlled decay

Low-resonance enclosure


I myself use almost exclusively PA drivers because they can deliver very high sound pressure levels and hardly distort at high hi-fi volumes—so for me, they meet one of the basic requirements for good speaker sound, and I feel quite at home in the SINAD forum.

I’m not arguing here about what’s subjective versus objective, because among the people here—most of whom I consider highly capable—there are at most misunderstandings in how things are phrased, and they’d likely all agree on what sounds good and clean if they were to get together for a blind comparison of different speakers. I don’t think anyone here has anything against measurements or a linear energy-frequency response.

With Sigberg, for example, I think that when it comes to the “finishing touches and final tuning” of his speakers, he mainly considers how a specific speaker performs in a specific room and tries to adapt his speakers to those conditions as best as possible. He also offers a calibration service for his speakers—that’s a truly useful service, which, by the way, is primarily supported by measurements.


And yes, I think it’s possible to design a speaker without listening to it at all. I’ve done it before; it was technically sound and sounded exactly as expected. But I also know pretty exactly what I want to have and achieve.

 

[SigbergAudio]

Another related and possibly interesting question that is more relevant for active speakers perhaps, is how aggressively you should correct a speaker. Going purely by the measurements, I should without a doubt employ 20+ high Q EQ points to the tweeter to make the on-axis as flat as possible. Will this however sound better or worse than three low Q EQ points (the measurement of the latter will look worse)?

 

[SigbergAudio]

I believe that many engineers at major speaker companies could design a driver from scratch—based on their experience—that would come very close to the ideal.
There are a great many companies, especially in the PA sector, that manufacture very good coaxial drivers. If you’d like, I’d be happy to name a few.

I don't understand this argument. Are you suggesting PA companies don't use measurements during development?

 

[amirm]

Why are there so many different approaches to loudspeaker design—leaving visual aesthetics aside—when it’s now so easy to achieve a flat on-axis frequency response using DSP and solid amplifier technology?

Because the a) the market for the most part doesn't want an active speaker and b) active speakers can have limitations with their amplification.

Otherwise, they are truly a superior solution.

As to why there are so many different approaches, it is because every random audiophile can get into making speakers with very little knowledge of electronics, etc. So many think that it is cool to be in audio business so jump in there with yet another design approach. Vast majority of the different speaker designs have no business being there as evidenced by so many more bad designs than good.

 

[amirm]

Another related and possibly interesting question that is more relevant for active speakers perhaps, is how aggressively you should correct a speaker. Going purely by the measurements, I should without a doubt employ 20+ high Q EQ points to the tweeter to make the on-axis as flat as possible.

Sounds either like a broken design, or you are trying to correct very small variations.

 

[Audionaut]

I don't understand this argument. Are you suggesting PA companies don't use measurements during development?

Thorbjørn, do you really think I would write such nonsense?
It seems to me that you’re getting very emotionally invested in this discussion right now and aren’t reading the questions accompanying the answer carefully.
Here it is:

Would anyone be able to build a driver from the ground up without measurements and end-up been decent?

 

[Audionaut]

As to why there are so many different approaches, it is because every random audiophile can get into making speakers with very little knowledge of electronics, etc.

That only addresses one small aspect of several questions.

Let’s get specific: You have the choice between a Neumann speaker, which you’ve measured and found to be very good, and a Genelec speaker, which you’ve measured and found to be very good. Both are similar in size and have a similar diaphragm area.

Which one do you choose, and why?

 

[TimVG]

That wouldn't tell us much as the issues may go beyond the crossover (think lack of waveguide).

I was actually just considering this as you responded. What if I'm asked to develop a crossover for a non-optimal system? Then it would probably come down to optimising for either the nearfield or the farfield, and things would become much more room dependent in general, so choices would have to be made. So perhaps this is where tuning by ear comes from: trying to balance out a system with flaws?

This is why I mentioned several pages back that these days even hobbyists can use software to predict with more than decent accuracy, potential system performance before as much as touching a sheet of MDF/ply.

A more relevant comparison would be two speakers at the same price point, one using measurements as the metric of performance, and the other, tuning by ear, and showing worse measurements.

Of course we can already predict the outcome

Semi-related anecdote; Spinorama.org provides EQ suggestions to improve(?) all speakers that are there. The improvements made favors on-axis response over off-axis response. These are adjustments I could have made, but would for instance result in excess build-up in energy off-axis at some frequencies. These are adjustments I've obviously tested along the way, but have discarded due to me making other choices, and finding them to sound subjectively better.

I could be wrong and biased of course, hence my reference to this as an anecdote.

It depends on what you're optimising for. I would assume your models are primarily aimed at living room (far field) listening. So unless directivity is perfect, a balance wil have to be made somewhere.

Not intended as criticism but a sincere question: have you considered moving towards non-coaxial models? There are some difficulties with most available coaxial drive units that I believe are making your process a bit harder than it should be. Just one thing I noticed but the coaxial unit you use has a distinct flare up which shows up in the off-axis behaviour, and is quite clear in the side-wall reflection average. Again not intended as criticism, but could it be that your tuning by ear process is simply trying to balance out these inherent little quirks?

 

[Sokel]

I believe that many engineers at major speaker companies could design a driver from scratch—based on their experience—that would come very close to the ideal.
There are a great many companies, especially in the PA sector, that manufacture very good coaxial drivers. If you’d like, I’d be happy to name a few.

You don't need to, I'll take your word for it even if the "famous" ones are few.
But that's not the question, the base of question was that, is it possible for we, to be so naive to think that AJ (or the companies you mention) can design a nice coaxial from the ground up without measurements?

He absolutely used measurements for sure, he even invented measuring systems at his early days at Kef,
The flowery nonsense at the interviews target the intended market.

We can argue all day about what's good for us, while a company (any company) would target at what sells and how.
And measurements alone cannot sell much, you need a story for this, marketing 101 (even if the story are these very measurements, depending the target) .

That's all.

 

[Sokel]

And about trends and stories, even about the objective "camp".

Just take a look at this little marvel here:

Small 2-way speakers with linear on-axis and power response characteristics (Scan Speak and SB Acoustics drivers). H&V off-axis measurements included

I've been trying to make design of a small, simple-to-build and inexpensive speaker that would not need a sub at least for music listening at moderate spl levels. No need for high power. This is third approach, code name "Mechano23". I'm sharing the results and VituixCad project file (LINK) with...

www.audiosciencereview.com

No trend, no waveguide other that the inherent Scan-Speak one, dump passive, etc.

Yet?

And designs like that are Monday at diyaudio.
No story though. No trend.

 

[Audionaut]

We can argue all day about what's good for us, while a company (any company) would target at what sells and how.

Yes, of course. But that only tells a small part of the story. Personally, I’m much more interested in the engineers and what they had in mind when designing them. I know some of them personally, and I find the developments and patents from BMS, JBL, and later B&C regarding ring-radiator compression drivers particularly fascinating.

I also have great respect for how Beyma maximized the BxL factor in the QLEX drivers despite the ferrite magnet, through clever coil winding.

I often wrote, talked, and discussed things with Josef Manger; the people at MBL had similar yet entirely different approaches—and what unites them all is that they were fully committed with their hearts and minds, far removed from the financial considerations that often hung over them like the sword of Damocles.

Cool idea—but we have to sell it somehow, too. Oh dear

I’m more into the cool idea and the people who came up with it—less into the selling.

 

[SigbergAudio]

Sounds either like a broken design, or you are trying to correct very small variations.

Yes, typically the latter. But we won't necessarily see if such a strategy has been employed from the measurements. So it is in part a constructed example to say that "maximum flat" isn't necessarily to be achieved by all means, and in part that there may be more information to be had than what we see in the measurements.

This is not an argument against measurements in any way, I am very much in the camp of them being important both for designing and evaluating speakers. But rather that it may sometimes be a bit more complicated to understand exactly what is going on.

 

[SigbergAudio]

Thorbjørn, do you really think I would write such nonsense?
It seems to me that you’re getting very emotionally invested in this discussion right now and aren’t reading the questions accompanying the answer carefully.
Here it is:

I can assure you I am perfectly calm.

I don't think you would, that's why I am asking to get clarification. But this didn't help. He is asking if anyone can build a decent driver without measurements. My reply would be a clear no.