Ideas for more meaningful speaker measurements

[markus]

Even though a Klippel is a very nice piece of equipment, it can only measure the speaker's performance. What do i mean by that? It measures how flat the frequency response is, we get a PIR, we get the DI, but what we don't get is what that does once we put it into our own homes.
Then there's the problem we have 2 ears we listen with, which means the left channel bleeds into our right ear and the right channel into our left ear.
While this may sound insignificant, it does cause dips and peaks in the frequency response we get at our listening position. Dips and peaks in a Stereo signal that fakes something (phantom center) that comes from right in front of us. If there was a sound source right in front of us, those dips and peaks wouldn't be there. A speaker that is uneven, right where those dips and peaks are, might actually sound quite convincing if those dips/peaks counter the cross talk effects somewhat, happening at the listener's ears. It might even be why some form of BBC dip still lives on while the actual reason for that dip was something different entirely (related to the speaker they were developing at the time). It has gained a new meaning for many and lived on as a tweak that can even have many meanings as to where that dip should be.
I know Floyd Toole mentioned that reflections will fill most of those (cross talk) dips quite well, but not without consequences. The timing of those reflections do matter, as it brings quite a different presentation if they are very early (<7ms) early (>7 ms <10) ms or late (> 10 ms or more). Most of the readers probably have not heard what those differences actually sound like. So how is one to know if they like a wide pattern or a more narrow pattern speaker. I guess (or hope) we can all agree that a smooth DI will help make a better speaker, but the pattern control needed is a lot more difficult without hearing the differences and knowing what it does in your room to/for your ears.
After following many discussions about this, where it is known that Toole advocates the beneficial nature of reflections while Dr. Earl Geddes tried to convince people we would have to avoid early reflections as much as possible, both of these theories might actually be (somewhat) right. Some consensus was found that those who like classical music tend to like a wide pattern more. Those who listen more to studio recordings like 'pop music' might prefer a more narrow pattern (or at least avoidance of early reflections (<10 ms).
Without knowing what one likes, the current reviews might not tell us enough. As it is the room's response to the chosen speaker within the room that determines a lot of how the sound and the presentation is going to be.
One cannot put a preference label on that as long as there's widely differing tastes and preferences out there. Aside from the previously discussed flaws in that system.

Just a few thoughts, after spending quite a long time trying to determine my own preferences.
(and finding ways to make use of the knowledge about cross talk and it's effects)
f.y.i. my preferences are avoiding/absorbing early reflections <15ms, while adding a Haas kicker at ~17ms.

Hope it fits within the discussion.

"Only"? I mean, that's the sole purpose of the damn thing. What you describe is a lack of scientific research in how we could relate metrics to perception. And I agree.

We still could answer a lot more useful questions with measurements alone:
- How flat is the response?
- How wide is the response?
- How does the "fingerprint" of the indirect sound look like?
- How much distortion does it produce with/without subwoofer?
- How loud does it go?
- How much "watts" does it need?
- "Something" that describes speaker-amp matching in a meaningful way

Could be a ranking system. Would help people in making more informed purchasing decisions. Would help us rise above those silly "my preference is better than your preference" debates.

 

[BenB]

For anyone who has gone through the Harman training, or otherwise knows how different frequency ranges are perceived, they can look at the plots we have and get a good idea of how well balanced the speaker will sound. (Unless that speaker has a very atypical dispersion pattern.) That's great, because spectral balance is typically the most important aspect of a speaker to get right.
Unfortunately, the next most important aspect for most people is soundstage/imaging, and we don't really have anything that addresses that directly. We know we need good matching between the speakers to achieve this, but even if we tested 2 speakers instead of 1, that's an insufficient sampling size to really get any feel for consistency, and the results would largely be random. We have a good idea that wide horizontal dispersion widens the soundstage, but we don't know to what extent, and we don't know what frequency ranges might be most important. (If someone has research on that, please point me to it.)
If there were any way to create a normalized apparent source width score for each speaker, I would be very much in favor of doing so. Similarly, if there were a way scientific (objective) way to score imaging, that would be extremely useful.
As speaker designs improve, and most companies are able to achieve a well balanced result, imaging / soundstage will be the biggest differentiator for consumers, and we have no way to help in that regard.

 

[markus]

For anyone who has gone through the Harman training, or otherwise knows how different frequency ranges are perceived, they can look at the plots we have and get a good idea of how well balanced the speaker will sound. (Unless that speaker has a very atypical dispersion pattern.) That's great, because spectral balance is typically the most important aspect of a speaker to get right.
Unfortunately, the next most important aspect for most people is soundstage/imaging, and we don't really have anything that addresses that directly. We know we need good matching between the speakers to achieve this, but even if we tested 2 speakers instead of 1, that's an insufficient sampling size to really get any feel for consistency, and the results would largely be random. We have a good idea that wide horizontal dispersion widens the soundstage, but we don't know to what extent, and we don't know what frequency ranges might be most important. (If someone has research on that, please point me to it.)
If there were any way to create a normalized apparent source width score for each speaker, I would be very much in favor of doing so. Similarly, if there were a way scientific (objective) way to score imaging, that would be extremely useful.
As speaker designs improve, and most companies are able to achieve a well balanced result, imaging / soundstage will be the biggest differentiator for consumers, and we have no way to help in that regard.

...or use narrow directivity speakers, reduce all reflections below threshold and use an upmixer to taste. That's what I'm doing – or trying to do. Wide(ish) dispersion speakers are and will always be just a band-aid for the intrinsic limitations of stereo.

“Habits die hard. The introduction of stereo in the 1950s gave us a left-to-right soundstage, but close-microphone methods, multitracking, and pan-potting did nothing for a sense of envelopment—of being there. The classical music repertoire generally set a higher standard, having the advantage of the reflectivity of a large performance space, but a pair of loudspeakers deployed at ±30° or less is not an optimum arrangement for generating strong perceptions of envelopment (as will be explained later, this needs additional sounds arriving from further to the sides). Perhaps that is why audiophiles have for decades experimented with loudspeakers having wider dispersion (to excite more listening room reflections), with electronic add-ons and more loudspeakers (to generate delayed sounds arriving from the sides and rear). In the control room, audio engineers now employ spatial simulation algorithms to break the monotony of a frontal soundstage by adding some low-correlation spatial sounds, even including binaural crosstalk-cancellation. All have been intended to contribute more of “something that was missing” from the stereo reproduction experience that, even at its best, serves only a single listener.
The real solution is more channels, giving the capability of delivering anything from a single point image through to spatial envelopment and immersion, as appropriate.”

Excerpt From
Sound Reproduction
Floyd E. Toole

 

[dannut]

...or use narrow directivity speakers, reduce all reflections below threshold and use an upmixer to taste. .....

Sadly both music upmixers, that can be 'seasoned to taste' are deprecated. (L7 and PL2 Music). Some lower-end receivers still can be had with PL2, but then you don't get decent room-correction with it (Dirac Live or Audyssey App) and no pre-outs.

 

[markus]

Sadly both music upmixers, that can be 'seasoned to taste' are deprecated. (L7 and PL2 Music). Some lower-end receivers still can be had with PL2, but then you don't get decent room-correction with it (Dirac Live or Audyssey App) and no pre-outs.

Those were replaced with DTS Neural:X, Dolby Surround, Auro-3D.

 

[dannut]

First 2 of those cannot be 'seasoned to taste' and are quite aggressive. Even 'centre spread' is missing on newer AVRs.

Last one is quite okay, but not quite an 'upmixer only'. Adds content-independent delayed signals (synthetic reflections) to different speakers. Also limited availability on AVR-s.

 

[BenB]

...
The real solution is more channels, giving the capability of delivering anything from a single point image through to spatial envelopment and immersion, as appropriate.”

I appreciate your input. However, my personal experience with every upmixer I've ever heard is that the sound is unnatural, and distracting in a way that stereo paired with wide horizontal dispersion is not. I'm not sure if the whole approach is flawed in some way that can never be overcome, or if all the implementations have been lacking, but this experience has made me very skeptical. I know a lot of others have had similar experiences, and remain similarly skeptical.

 

[markus]

First 2 of those cannot be 'seasoned to taste' and are quite aggressive. Even 'centre spread' is missing on newer AVRs.

True.

Last one is quite okay, but not quite an 'upmixer only'. Adds content-independent delayed signals (synthetic reflections) to different speakers. Also limited availability on AVR-s.

Putting a speaker in a room also adds "content-independent delayed signals" aka reflections

 

[markus]

I appreciate your input. However, my personal experience with every upmixer I've ever heard is that the sound is unnatural, and distracting in a way that stereo paired with wide horizontal dispersion is not. I'm not sure if the whole approach is flawed in some way that can never be overcome, or if all the implementations have been lacking, but this experience has made me very skeptical. I know a lot of others have had similar experiences, and remain similarly skeptical.

Depends on preference and expectations. I think that many have "adapted" to stereo so they find anything else "unnatural". "Stereo sound" is certainly more a cultural development than a technical one so it could be argued that it "needs" to sound as it sounds and anything that adds spaciousness and envelopment to a recording makes it sound "unnatural".

 

[Wesayso]

However, my personal experience with every upmixer I've ever heard is that the sound is unnatural, and distracting in a way that stereo paired with wide horizontal dispersion is not.

Not quite how my experience is regarding that last part: the wide dispersion.

Wide dispersion sure adds a widening to the stage/imaging but it's the 'same sauce on everything' that grows old quickly (for me anyway).
It isn't adjusting to what's in the recording. It sort of adds the same effect to everything.

That's reason enough for me to absorb the early reflections and add later 'virtual' queues.
I do my own mixing for the virtual part and it has worked quite well, when used with some care (SPL level and timing).
One of the key ingredients is to have it decorrelated from the main program. I did try some of the up-mixers mentioned
but wasn't convinced by the results. Sadly I haven't been able to test L7 or Auro-3D (yet).

 

[DSJR]

Distortion apart, how can one differentiate a 'port-bound' underdamped boomy speaker from a well extended but perceived 'tauter' sounding model. S'cuse clumsy wording, but 'boom' doesn't always seem to show in response measurements, although it's obvious apparently what it'd sound like from knowing the characteristics of the drive unit, the box it's in and port tuning and so on... is measured distortion the only way as shown?

 

[markus]

Distortion apart, how can one differentiate a 'port-bound' underdamped boomy speaker from a well extended but perceived 'tauter' sounding model. S'cuse clumsy wording, but 'boom' doesn't always seem to show in response measurements, although it's obvious apparently what it'd sound like from knowing the characteristics of the drive unit, the box it's in and port tuning and so on... is measured distortion the only way as shown?

Anechoic group delay will show it.

 

[dc655321]

One of the key ingredients is to have it decorrelated from the main program.

Curious, along what dimensions are you decorrelating? Time, amplitude, etc

 

[JRS]

That's why I think score without sub is almost meaningless and there is score with subwoofer for that.

For anyone who has gone through the Harman training, or otherwise knows how different frequency ranges are perceived, they can look at the plots we have and get a good idea of how well balanced the speaker will sound. (Unless that speaker has a very atypical dispersion pattern.) That's great, because spectral balance is typically the most important aspect of a speaker to get right.
Unfortunately, the next most important aspect for most people is soundstage/imaging, and we don't really have anything that addresses that directly. We know we need good matching between the speakers to achieve this, but even if we tested 2 speakers instead of 1, that's an insufficient sampling size to really get any feel for consistency, and the results would largely be random. We have a good idea that wide horizontal dispersion widens the soundstage, but we don't know to what extent, and we don't know what frequency ranges might be most important. (If someone has research on that, please point me to it.)
If there were any way to create a normalized apparent source width score for each speaker, I would be very much in favor of doing so. Similarly, if there were a way scientific (objective) way to score imaging, that would be extremely useful.
As speaker designs improve, and most companies are able to achieve a well balanced result, imaging / soundstage will be the biggest differentiator for consumers, and we have no way to help in that regard.

I should think precision of imaging is testable. It would require specific recording that moves an instrument &/or vocalist around in the x plane and the listener uses some pointing device to indicate where in space the sound seems to originate including boundaries indicating how diffuse the image is.

The hard part is determining the specific angles and distance the center of the speakers make with the listener that is needed to optimize localization. Be nice to have the speakers on a horizontal slide and a turntable, controlling both angle wrt to the listener and toe in to be varied. If needed an additional track for the listening chair enables the listener to move back and forth wrt to the speakers to determine optimal distance away from the speakers. Then we can sort out the y plane.

Hell I'd love to have one of these in my own home. This all somewhat tongue-in-cheek, just playing with the idea, and based a bit on the Harman labs set-up. Could definitely sort out near field vs far field and give some idea how clearly images are rendered. Of course the timbre may also change as the speaker is toed in and out, but that's where a stable DI should help.

 

[Wesayso]

Curious, along what dimensions are you decorrelating? Time, amplitude, etc

Mostly time, as in no direct sound from ambient channels but let them diffract/reflect of nearby objects plus adding some subtle reverb.
The tonal balance still is of influence, thus important to preserve. My main inspiration comes from the work of David Griesinger. He has many papers on this subject (on his website) and has written reverb algorithm's (Lexicon) and was the driving force behind Harman's Logic 7.

I've also experimented with using IR's of real room/space reverbs but the above turned out to be much more versatile.

 

[storing]

Maybe this will help. Einstein understood the general relativity but he was unable to test it. He did though formulate it. Scientist that followed him used his formulas and finally proved it in tests.

In science, everything starts with a concept, followed by mathematical description then test to prove it. In engineering (and some sections of medicine) this may not be the case as scientists already gave us the formulas. We can start with the tests.

I understand your example, but still I find it strange you use it to describe science as a whole. It's just one very specific example from physics, and yes it happens to be the concept->formulate->verify principle, but that really doesn't extrapolate to any (even narrow) definition of science I ever came across. Could be difference in our backgrounds, or language, or perhaps you mean to describe physical sciences only (but that still wouldn't cut it, there's enough chemistry/geology/... out there which does not follow that principle).

 

[Holmz]

Mostly time, as in no direct sound from ambient channels but let them diffract/reflect of nearby objects plus adding some subtle reverb.
The tonal balance still is of influence, thus important to preserve. My main inspiration comes from the work of David Griesinger. He has many papers on this subject (on his website) and has written reverb algorithm's (Lexicon) and was the driving force behind Harman's Logic 7.

I've also experimented with using IR's of real room/space reverbs but the above turned out to be much more versatile.

It is easier to just do a cross-correlation with the input signal, and the reflections all just jump out like a comb.
The low freq stuff doesn’t really show up there though.

 

[Wesayso]

It is easier to just do a cross-correlation with the input signal, and the reflections all just jump out like a comb.
The low freq stuff doesn’t really show up there though.

Reflections aren't a bad thing if the goal is to emulate a bigger/better room. See the combing in a real room.
DSP takes care of the perceived balance. As long as nothing sticks out this is quite doable.

 

[sarumbear]

I understand your example, but still I find it strange you use it to describe science as a whole. It's just one very specific example from physics, and yes it happens to be the concept->formulate->verify principle, but that really doesn't extrapolate to any (even narrow) definition of science I ever came across. Could be difference in our backgrounds, or language, or perhaps you mean to describe physical sciences only (but that still wouldn't cut it, there's enough chemistry/geology/... out there which does not follow that principle).

I gave you an example. I cannot even comprehend science as a whole, how on earth I will try to explain it?

 

[DanielT]

I gave you an example. I cannot even comprehend science as a whole, how on earth I will try to explain it?

OT

For those who are interested in science:

Ha ha, she says 3:35 into the video to the economics award winner: You're not a scientist