Can tweeter detail capability be rated objectively?

[whitwye]

I collect guitars as well as speakers, and I can reasonably well hear the difference between speaker accuracy -- or at least plausibility -- in reproducing the harmonics of a guitar. It's one thing to get a clear representation of the fundamentals, of the basic note played. It's quite another to get the shimmer of the overtones. What puzzles me is that in speaker marketing there's never a reference to any sort of testing for that capability of reproducing the complexity of a sound. As I've been cycling through different speakers in my main room, mixing brands variously, the comparisons have been obvious. Whatever is in the Fluance towers I got years back (and cheaply), not much good; Klipsch bookshelf speakers -- several models -- bell like fundamentals, but metallic distortion of the overtones; Polk bookshelf and center (lower end of their lines), inferior even to Klipsches, if not as harsh; Emotiva T1s, immensely detailed; RSL CG5s, equal to (and compatible with, and priced like) the T1s; RSL CG3s and 23, not quite the equal, but blend quite nicely -- and far beyond similarly priced Klipsches and Polks.

This is all frequencies and waveforms. There have to be objective tests available that could capture these differences. Has anyone devised them.

[garbulky]

The closest I can think is Audio Science Review. They measure the speaker which includes contour plots which show the radiation patterns from the tweeter versus the woofer etc. Then they do a listening test and try to describe what is heard.
For instance, Emotiva Airmotiv 6 s review:
www.audiosciencereview.com/forum/index.php?threads/emotiva-airmotiv-6s-powered-speaker-review.11185/
I personally think I would have been far more impressed subjectively with the airmotiv 6 than the review was, but at least he gave it a go.

[DYohn]

ASR is often wrong, in my experience.

Measuring the characteristics the OP describes is almost impossible, especially since much of what he is describing is perceptual and you simply cannot measure that. You have to listen.

[mfeust]

I think some of the measuring and testing that Harman Audio group and Dr Floyd Toole are doing would probably be the closest.

Mark

[Boomzilla]

On a test bench, tweeters CAN be measured accurately. Time to rise - time to return to zero after an impulse - ringing - overshoot - various types of distortion - etc. But these measurements still don't tell the whole story (measurements NEVER do). Why? Because the characteristics of the tweeter change RADICALLY with volume increase. How perfectly a tweeter performs at whisper volumes tells you nothing about how it performs at higher volumes. Now on the face of it, this would seem like a minor impediment - just measure the tweeter at multiple volumes. And although the multiple-volumes test might allow you to graph the volume-dependent performance of the tweeter, the information will still be useless.

Ignoring, for the moment, the fact that there can be fairly large manufacturing variations in the performance of some tweeters, and that over time every tweeter's performance will change (particularly true of some tweeters that use ferrofluid damping), the core impediment to measuring tweeter performance is music itself.

Music is not a series of subsequent pure tones. It has multiple tones simultaneously. Tweeters deal differently with how they respond to these simultaneous tones. The music also doesn't play at a single volume. Loudness is constantly changing. Add to these variables that the listener sometimes plays music louder or softer depending on a variety of variables that the speaker designer has no knowledge of or control over.

The combined result is that a tweeter that performs admirably in bench testing at a specific frequency and at a specific volume may sound like dog droppings in the real world. No tweeter is consistent throughout its performance range. Dr. Floyd Toole's experiments, to my understanding, are primarily concerned with how dispersion affects musical perception. But a tweeter with excellent dispersion can still perform poorly.

And these issues, in a nutshell, are why I agree with Mr. DYohn on this particular question. Measurements, unless (until?) we discover new ways to measure, will never be able to quantify tweeter performance beyond the issues mentioned in the first paragraph of this post. And I must also opine that the OP is wrong when he contends that there is consistency over different models of a single brand. It's wrong to say that "Klipsch tweeters have metallic distortion on the overtones" when the company uses a wide variety of different tweeters in their products. Unless you have heard ALL the different tweeters that the brand uses, I don't think you can lump them all together. It's also wrong to characterize tweeters by design characteristics - I've heard soft-dome tweeters that sounded bad, and others that I thought could compete with electrostatic panels. You just have to listen. Really.

[audiobill]

Well said, Boom.

[405x5]

www.stereophile.com/content/glorious-time-ars-edgar-villchur-and-roy-allison-allison-part-2

[KeithL]

A good set of waterfall plots will tell you quite a bit about how a tweeter will sound...

(Of course you have to know how to interpret them.)

A lot of folks seem to like to use waterfall plots for things like measuring bass traps...

But they can also tell you a lot about a tweeter...

(But, in order to be good for that, you need to start with a really clean microphone, and proper test signals.)

[Boomzilla]

Jun 29, 2020 11:55:22 GMT -5 KeithL said:

A good set of waterfall plots will tell you quite a bit about how a tweeter will sound... Of course you have to know how to interpret them...

My objection to waterfall plots (if I understand them correctly) is that although they look at a sound source over a range of frequencies, they are performed at fixed volumes. What the waterfall looks like at a lower volume is not necessarily what it would look like at a higher volume. But I may misunderstand?

[klinemj]

I've always said that I'm sure someone could figure out how to actually analyze sounds in technical measures and correlate it well to human perception, however it would be horrendously complicated and likely in the form of a model incorporating many factors (like those Boom and others have noted). And, of course, there's always George Box's point that all models are wrong and some are useful...so "ya got that goin for ya'..."

That said, in my career, I have seen some very impressive sensory modeling that was very useful. A friend correlated 7 lab measures to consumer ratings on a particular sensory variable (can't tell you what, but it related to the sense of touch but also had to include visual aspects also because they were huge factors). I thought he was going after a folly when he told me his plan, but it worked! And, we successfully used his model for years, then adapted it as we created new technologies that went beyond the model's inputs.

P&G has many experts in sensory analysis, and they have impressed me with how well they can model complex multi-variable systems.

And...that's really the challenge for modeling consumer perception of sound. First, there are many technical variables, and second - human response is not uniform either for ANY sensory metric.

Mark

[MusicHead]

Although it does not address directly the specific question from the OP, I find this talk by Dr. Floyd Toole quite interesting. It was given in 2015 at McGill University:



You may or may not agree with his conclusions and methodology (or even the premises on which both are based), but I believe what he has to say deserves consideration. 

The video is fairly long, about 1h15m. If you watch it, do not go skipping around. Single portions taken out of context will definitely give you a distorted view of what he wanted to communicate to the audience.

[whitwye]

Thanks for all the informative responses. Yes, musical tones, particularly overtones and chords, are complex. There's also the issue of tonal coherence over time -- is what coloration is there consistent, such that the brain can tune it out, as we do with the variation in frequencies of sunlight over the span of a day. Yet it would seem that in principle, for a given instant's cross section of a musical recording, that it should be possible to graph what's there -- the entire set of frequencies and amplitudes. Then given a near-perfect microphone in a neutral chamber, it should be possible to graph what a speaker puts out, and see what the correlations are between the two graphs.

Something I've discovered by slowly evolving my now 5.2.1 (or 7.1, depending on source) set of speakers is that while any combination sounds reasonably good for a time, over weeks of listening the ears learn which are the clearest, most real. And the less clear become less tolerable. I've had no such discord in the current set now: T1 L & R, RSL CG5 heights, CG23 center & CG3s in back. Yes, the smaller center and rears have an obvious smallness, but not a discordant one. The last iteration had Klipsch RB-51s as heights, and R-41m's in back, and their lack compared to the T1s got to where it was nearly painful. The G3s are priced midway bewteen those Klipsches, yet far better. Some Polk TSI-100s from a prior iteration still serve well in another room, but were far inferior in the mix even to the Klipsches, even the similarly priced R-41s. As for the metallic limitations of Klipsch horns, discussion elsewhere claims the problem was in the past, so should have been present in the RB-51s but not the newer R-41s. That's not how I found it -- although for a bell-tone or flute, Klipsches are excellent.

This experience of awareness of quality difference emerging from prolonged listening makes me wonder if if there's something to doing A + B testing -- both at once -- rather than the normal A / B testing -- one at a time. Listening to speakers combined, over time the ears sort them out, even where the initial reaction, or even A / B comparison, finds no such strong favorite. I'm guessing part of this comes down to phase effects between their outputs that would be very hard to pin down on a graph though. Still, there might at least be some signature of speaker compatibility which would emerge.

Lest my going on about the RSLs seem a bit disloyal here, one difference I am seeing between the otherwise equally excellent T1s and the CG5s is that with the T1s being on my A-300, and the CG5s on my A-700, at higher volume in all-channel stereo (balanced forward) I can hear what I take to be the A-700s lower headroom. Wednesday I should have a couple of Stealth PA-1s to hopefully fix this.

[vcautokid]

After all this. You are still going to have to listen to it to hear if it what you like. And that you cannot measure. As emotion is not quantifiable. That is why the science is that. But I got to still like it in the end. The time where the end is as important as the journey.

[405x5]

Jun 27, 2020 14:49:32 GMT -5 whitwye said:

I collect guitars as well as speakers, and I can reasonably well hear the difference between speaker accuracy -- or at least plausibility -- in reproducing the harmonics of a guitar. It's one thing to get a clear representation of the fundamentals, of the basic note played. It's quite another to get the shimmer of the overtones. What puzzles me is that in speaker marketing there's never a reference to any sort of testing for that capability of reproducing the complexity of a sound. As I've been cycling through different speakers in my main room, mixing brands variously, the comparisons have been obvious. Whatever is in the Fluance towers I got years back (and cheaply), not much good; Klipsch bookshelf speakers -- several models -- bell like fundamentals, but metallic distortion of the overtones; Polk bookshelf and center (lower end of their lines), inferior even to Klipsches, if not as harsh; Emotiva T1s, immensely detailed; RSL CG5s, equal to (and compatible with, and priced like) the T1s; RSL CG3s and 23, not quite the equal, but blend quite nicely -- and far beyond similarly priced Klipsches and Polks.

This is all frequencies and waveforms. There have to be objective tests available that could capture these differences. Has anyone devised them.

You’re right on track, doing what the rest of us have been doing for decades. (Swapping and evaluating by ear systems to get as close as possible)
Some of what you’re looking for has been available from a few audio engineers, but the vast majority of end users go by what they hear and are not going to care how they got there or micro manage it in the process.

Bill

[Deleted]

405x5 says:    the vast majority of end users go by what they hear 

DYohn says:     You have to listen

Boom says:     You just have to listen. Really.

Vince says:      listen to it to hear if it what you like

................................................................................................................................


Yes, I initially go objective by FR graphs, specs and other measured info.  However, in the end it is mainly subjective for my purchase decision.

I use a magnificent test machine, my EARS!



Floyd Toole ..... 1 hr - 15 mins ..... Sorry, I'm sort of busy right now watching the Seahawks/Eagles game, 
 
           

[pedrocols]

My speakers measure wonderfully all the way up to 20khz. However, my ears don't measure that great. They start to drop drastically at about 13 to 14khz.

[Deleted]

Jun 27, 2020 14:49:32 GMT -5 whitwye said:

I collect guitars as well as speakers, and I can reasonably well hear the difference between speaker accuracy -- or at least plausibility -- in reproducing the harmonics of a guitar. It's one thing to get a clear representation of the fundamentals, of the basic note played. It's quite another to get the shimmer of the overtones. What puzzles me is that in speaker marketing there's never a reference to any sort of testing for that capability of reproducing the complexity of a sound.

This is exactly what Tekton Design attempts to accomplish with their patented MTM array [low mass algorithm]. Check out the video 5:05 into it.