What's Left In Speaker Design To Reduce Distortion/Increase Detail Retrieval?

[tuga]

Why do you think I used the phrase off-the-shelf drivers?

I think I should stop this chat as it’s obvious I couldn’t explain what I mean to you and it seems to me you are arguing for the sake of arguing.

There are several manufacturers making compression drivers (BMS, JBL , RCF, B&C, Celestion - I recall watching a talk by Kef's @jackocleebrown about a new Celestion model, I think it was a compression driver). Yes they're mainly directed at the PA market but they are available. And I don't doubt that a huge multi-way horn is a tough sell when you can get an elegant and slim R9 for a lot less money.

 

[Oristo]

I like the way that tall narrow ribbon speakers can sound;
as line sources, their lateral dispersion is good.
Since their entire radiating surface is driven, coherence is possible.
SPL drops off more nearly proportional to distance than square of distance.
Weakness is lack of bass and problematic integration with woofers.

 

[goat76]

I don't see them as the same thing. Passing the file through a medium, electronic or acoustic, should yield the same signal if accuracy to the file is the goal. For a single speaker, this is achievable in theory, but not when going stereo.

The second accuracy goal - to sound identical to the final product in the studio. Then we should copy both on-axis frequency response and the DR ratio. If we can copy on-axis frequency response (and playing at a given reference SPL to take care of FL-M curve), we would not improve accuracy if we e.g. use a speaker setup with higher DR ratio than that used in the studio.

People can set up their stereo speakers in all kinds of funny ways, and they can have an awful-sounding room caused by really bad acoustics, so we should probably not even consider those types of faults because that will bring a lot of outside factors into the game.

If we instead follow the studio standard of an equilateral stereo setup in a fairly good-sounding and acoustically treated room, we will at least end up "in the same ballpark" of how that stereo mix was intended to sound in a stereo setup. That was the main goal and the initial aim for the final product, so how the stereo mix sounds on a single speaker in mono can in this case never be seen as the accurate version or the most accurate sound. The stereo mix may have been briefly checked for mono compatibility, that's probably all when it comes to mono listening because the stereo mix was first and foremost intended to be listened to in stereo.



So what is the right way to go for us music consumers?

We can hardly know how the mix sounded in the mixing studio, and we don't know if the file would have ended up differently if the mixing engineer had been using another studio but with the same sounding goal. All we know is that the file ended up as it did because of a bunch of unknown factors, and it's possible that even the mixing engineer himself wouldn't like the mix if he heard it in an "accurate" sound system. What ended up in the file is the ”fruit” of what the mixing engineer heard, you can't separate those two things.

So what does that lead to?

We can choose the speakers we find sounding the most accurate to ourselves or go for the objectively best measuring speakers and take it from there. We can EQ the system to our liking so that it suits most of the music we usually listen to, and/or we can choose to EQ all the records themselves until they suit our own liking. But we can never be completely sure what is really accurate when it comes to the program material, or if that was even the goal of the creator of the content.

"Circle of confusion"

 

[goat76]

In your example it’s not the drivers but the horn that increased the efficiency. Drivers are standard, off-the-shelf units. You increase efficiency slightly but create a speaker which is challenging to use in a domestic room. What is needed is a complete paradigm change where the efficiency is increased by order of magnitude.

Imagine the audio system when average efficiency is 130dB/W or more. Everything we take it for granted will change, everything! Why would we need a power amplifier when a headphone output can power any speaker to deafening levels.

That is what I’m talking about.

I know, this video has an intentionally provocative title, but Klaus Heinz from HEDD Audio is talking about the same thing you do, and other interesting things.

 

[Axo1989]

The Harman shufflers are a brilliant idea, same piece of music, level matched within four seconds, I really can’t understand your continuing ridicule of them.
Keith

Pointing out that 'same position in room and no toe-in' for a stereo pair speaker comparison has some shortcomings isn't the same as ridicule. Assuming that's what the set-up was (I think there were/are a few different setups using the shuffler method).

 

[MattHooper]

How can you evaluate the stereo effect with a single speaker?

By correlating the relationship between the sound of single speakers and the sound of stereo speakers. If there's a good correlation, then mono attributes predict stereo attributes.

 

[Axo1989]

I think I should stop this chat as it’s obvious I couldn’t explain what I mean to you and it seems to me you are arguing for the sake of arguing.

You've broken my irony meter. Can I PM you the repair quote?

Edit: not to dismiss your point here, which is interesting. Not the same thing exactly but how does IEM drive efficiency stack up?

 

[Ron Texas]

Secret sauce will do it.

 

[Blumlein 88]

I always appreciate your contributions!

Just riffing off what you guys are debating...

1. The discussion of how much room reflection we might want to hear in regard to speaker dispersion somewhat relates to my general question: Taking stereo recordings as the example, and presuming we are talking about the "best" performing current speakers: In terms of accurately transcribing the sonic information in to sound, so the listener can in principle hear all the sonic details contained in recordings, Are We There Yet? Is (well designed) speaker distortion now so low, the sound so accurate to the signal, that no sonic information is being lost at all? If so it suggests the question simply moves to how do we want that detail presented. E.g. with more spaciousness (for instance availing ourselves of room reflections), less spaciousness etc. In other words, now it's all about choices for dispersion/directivity characteristics?

However, if even the best loudspeakers are still seen as significant enough source of distortion that could be masking some recorded detail, where are the areas that are problematic that we'd want to improve? (I appreciate that people are giving a range of opinions).

I think the issue of distortion in speaker playback needs more work. I think Harman's work pushed them toward the idea it was of a lesser significance than the things they use in how they design speakers. Is it of no consequence in advanced designs? I don't know.

As an anecdotal touch point, consider the small LSR speakers from JBL. They can sound surprisingly good. One thing I note is used in a larger space than they are intended for they can get to sounding what I'd describe as opaque at higher volumes/peaks. In Amir's listening of the LSR 308 he noted how a compressor/limiter built into them caused various other sounds and distortion when he tried the 96 db testing. Of course a limiter is much higher and a different kind (lots of harmonics) of distortion than you'd get with a speaker driver unless it is ready to explode. I find inexpensive Revels not to have this opaqueness at higher volumes. I would think the limiter is to protect the amps and let the speaker have some use at higher levels. When used as a monitor you'll likely not reach levels like this anyway. So a reasonable trade off for the purpose.

Testing indicates with most music you'll need 3% to hear it and probably with no music do people pick up less than 1% distortion. Speakers that stay mostly below 1 % and not more than 3% on peaks are likely not to intrude with distortion. You do see in using REW or in Amir's tests that many speakers do have some narrow areas they test or exceed these levels. Most are up there in the lower 200 hz, though people apparently aren't as sensitive at those frequencies to distortion. The larger Revels above 200 hz mostly show one or two tenths of a percent distortion at Amir's 96 db test level. I doubt that much distortion is an issue at all. Larger Genelecs have these low levels of distortion as well. Ditto for the larger Neumann's and KEFs. So I'd think at least with larger speakers distortion is a non-issue. With smaller speakers maybe not.

2. Mono vs Stereo speaker blind listening tests: I wouldn't dispute Toole's research or your characterization. Though it's interesting to ponder the implications nonetheless.
As I've said before: If the main point of testing in mono was that stereo made it harder to identify sonic detriments in the sound, essentially masking artifacts that require mono to hear...then the relevance to stereo listening seems eroded. The analogy I've made before is like saying that you really have to taste wagu beef on it's own to discern how much better it is than cheap off the shelf ground beef. Putting it in your chili recipe will make the superiority of the wagu impossible to discern. But then, IF the point of buying beef is that you'll be using it for chili, it doesn't make sense to worry about buying wagu beef. Buy the cheaper beef, or whatever beef you like, since it will taste similar. One could say the same for buying/evaluating speakers. If the use case is listening in stereo, "why not evaluate speakers in stereo if that's how you'll be listening?" And if "worse quality" pair whose artifacts would have shown up in mono sounds perfectly fine in stereo....what's the prob?

That isn't to say a coherent argument for the relevance of mono-evaluation to stereo listening can't be made. As I recall, Toole has pointed out that among the justifications for mono tests is that some studies indicated that preferences in mono listening tended to predict very well preferences in stereo evaluations. Which is good. If so, that may indicate mono listening is *sufficient* for evaluating stereo performance. But it wouldn't establish it as *necessary* or even "better" for the purpose of sonic preferences in stereo. So long as a wider array of speakers would pass the stereo listening tests for "good sound" (because some artifacts will be less apparent in stereo), in principle this opens up the field of speakers that will sound "good" to the consumer, rather than just sticking to those that sound good in mono.

Well remember, in the stereo vs mono testing there were still discernible differences. It is not like all the pretty good in mono speakers became indistinguishable in stereo. Only that the level of preference was over a smaller range than in mono. So mono testing predicts good stereo performance, does so more clearly and makes stereo testing for stereo not informative on any level. Seems the speakers had not reached a level where some speakers were detected as better in mono, but found equally preferable with lower scoring speakers when tested in stereo. The other thing I think of is the very high correlation in their model versus bookshelf speakers or speakers with little output below 100 hz. It was nearly perfect. With larger speakers less so. One would think this is because of room effects at lower frequencies. One might imagine stereo speakers exciting the room differently than a central mono speaker. Which might alter results a bit. Maybe an area that needs more work, but given what is known about uneven low end response not really surprising.

3. Blind Speaker Tests in general: The blinded speaker research so often cited here seems clearly sound and informative. Informative in terms of establishing, in blind tests, strictly on sonic attributes, what people tend to prefer. However, again, pondering the implications is interesting. As I've argued before, there is a conundrum, perhaps even a paradox, lurking in the relevance of the research for consumers. It seems mostly just assumed that the result of the blinded conditions are relevant to the sighted conditions in which consumers will actually be listening to speakers. It's generally asserted as helping guide our speaker purchases. Yet...as far as I'm aware, there is little if any data showing how preferences established in blind listening tests predict preferences in sighted listening conditions. It's just sort of assumed there is some relevance. (And insofar as we are presented results of sighted vs blind tests, it is typically used to emphasize the differences, the unreliability of sighted listening, rather than the reverse: searching for how the blind test results might share any through-line trends with sighted test results).

The problem inherent in this situation is that: If such studies demonstrated that what is perceived in the blind listening tests predicted what is perceived in the sighted listening results, that necessarily erodes the relevance of the blind listening tests. It would suggest the relevant "good sound" characteristics can be evaluated in sighted listening too. But to the degree such studies would fail to predict sighted listening preferences from the blind test results...that too would erode the relevance of the blind tests to the actual use-case scenario for speakers. So we have this apparent gap...right where you'd want the research to be demonstrated as relevant!

Given this gap, in terms of the relevant to our preferences for sound, what kind of rational is there for choosing the speakers that measure in ways that people select in blind testing? It would seem the best one could say is that it satisfies someone's intellectual itch, that they at least know "this is rigorously determined as the best sound in blind tests." So in a way it's like specs-chasing SINAD, where one chooses the best SINAD (or other distortion metric) measuring component, from among others that are already below audible distortion, just for a sense of peace of mind or intellectual satisfaction. But moving beyond that to anything like "I've chosen this because I will perceive the sound as better under my sighted listening conditions" seems to bring on the issues cited above.

(As I've said, this is not so much a problem if we take sighted listening to be at least accurate enough, perhaps over time, to what we'd hear in blinded conditions).

The tests indicated a near disconnect from sound quality and speaker preference when sighted. Sighted bias is going to be the most with speakers because of how different they look. A separate path of research might be to create speakers with near equivalent performance with many different appearances and see if you can create designs that reinforce actual performance by biasing the sighted listener to expect great sound. Even that might prove difficult as much of sighted bias is knowing the cost and some story behind the brand. I mean what would you expect if an identical speaker to a big Wilson was called the Gene Eric Super Sound model II offering SOTA sound for $1500/pr instead of knowing it was a more than 10 times as expensive speaker with this super high tech material for the enclosure? Just the enclosure in one case will get called a cheap plastic like composite while in the other it is a high tech material with damping properties designed for good sound. And of course the price.

 

[Axo1989]

... So mono testing predicts good stereo performance, does so more clearly ...

Yes.

... and makes stereo testing for stereo not informative on any level. ...

And no.

 

[Bjorn]

Accuracy to the recording or accuracy to what is heard in the studio? If it is accuracy to the recording, the stereo flaws creates a problem, and that especially when DR ratio is high. If it is accuracy to what is heard in the studio, we need to recreate the same DR ratio all other things equal. Possible but quite difficult.

True to the recording. Removing specular reflections gives a higher resolution into the recorded signal. Hey, if you don't believe so, test it for yourself! Set up some broadband absorption panels at both side wall reflections (measure to make sure they attenuated well) and listen with and without to different music.

 

[Bjorn]

≥ 4-way horn speakers can achieve high sensitivity and constant directivity, for example this one.

That's not a constant direcitivty horn FIY.
But sarumbear is right. Even the highest sensitive speakers we have are literally insensitive where most watt is turned into heat.

 

[tuga]

By correlating the relationship between the sound of single speakers and the sound of stereo speakers. If there's a good correlation, then mono attributes predict stereo attributes.

Not really because differently beam-width will interact differently with the side walls and you cannot test that with a single speaker placed in the middle of the room.

 

[tuga]

That's not a constant direcitivty horn FIY.
But sarumbear is right. Even the highest sensitive speakers we have are literally insensitive where most watt is turned into heat.

It is omni only in the sub- and low-bass.
I can't find measurements for one of the multi-way Cessaros or Avangardes but these is from the smaller Avangardes:

Avantgarde Acoustic Uno Nano loudspeaker Measurements

Sidebar 3: Measurements

www.stereophile.com

 

[Purité Audio]

I used to sell ( or try to) the Cessaro range these were the ‘Liszt’ four way, and next to them some Avantgarde Zeros, it was hearing the then new Gen 8361s /7360 that made me realise how odd the horns sounded like listening to three different loudspeakers and that was from five metres.



Keith

 

[Thomas_A]

True to the recording. Removing specular reflections gives a higher resolution into the recorded signal. Hey, if you don't believe so, test it for yourself! Set up some broadband absorption panels at both side wall reflections (measure to make sure they attenuated well) and listen with and without to different music.

I’ve done experiments in the past both damping, near vs farfield etc. DR ratio need to be kept in control but also balance it to the stereo errors. They are audible as well so there is no ideal solution. Just compromises.

 

[Bjorn]

It is omni only in the sub- and low-bass.
I can't find measurements for one of the multi-way Cessaros or Avangardes but these is from the smaller Avangardes:

Avantgarde Acoustic Uno Nano loudspeaker Measurements

Sidebar 3: Measurements

www.stereophile.com

Vertically such horns will be a complete mess with serious lobing and comb filtering besides diffraction issues. Would need to see proper measurements of the horizontal.

 

[tuga]

Vertically such horns will be a complete mess with serious lobing and comb filtering besides diffraction issues. Would need to see proper measurements of the horizontal.

I agree but but it's a narrow-beamwidth mess, and maybe not as messy as you expect.
On the plus side the drivers are operating in the horn's optimal acoustic range do no massive EQ is required.

Avantgarde Acoustic Uno XD Messungen - FIDELITY online

Die Avantgarde Acoustic UNO XD glänzt im Messlabor durchweg mit beeindruckenden Messwerten, die den positiven Höreindruck massiv unterstreichen.

www.fidelity-online.de

 

[Bjorn]

I’ve done experiments in the past both damping, near vs farfield etc. DR ratio need to be kept in control but also balance it to the stereo errors. They are audible as well so there is no ideal solution. Just compromises.

We are not talking about personal preferences here. We are talking about how to retrieve most details. Both specular reflections and resonances have a major impact on this. It's not a coincidence that almost all studios create a reflection free zone within a certain time gap. It's crucial in order for them to hear the recorded signal as accurate as possible. And the same goes for playback of the recording. Whether one prefers something else, it's a different topic.

 

[tuga]

We are not talking about personal preferences here. We are talking about how to retrieve most details. Both specular reflections and resonances have a major impact on this. It's not a coincidence that almost all studios create a reflection free zone within a certain time gap. It's crucial in order for them to hear the recorded signal as accurate as possible. And the same goes for playback of the recording. Whether one prefers something else, it's a different topic.

Horns (proper, multi-way) seem to be a good way to retrieve more detail due to their increased sensitivity and narrow directivity.

I think that the dislike for horns, apart from price and aesthetics, is probably due to the reduced interaction with the room. The listener is 'spot-lit' with the recording and some people don't enjoy the sharp imaging or the reduced sense of 'envelopment'.

I miss the 'effortlessness' I got from my half-hacked horn speakers.