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

[Blumlein 88]

Thank you so much. Do you happen to know what that directivity was, or name/citation of the paper(s) that was published in?

Best source is Toole's book. You can also read several blogs by Sean Olive.

 

[NTK]

Thank you so much. Do you happen to know what that directivity was, or name/citation of the paper(s) that was published in?

If you are referring the directivity index curves (sound power directivity index, SPDI, and early reflection directivity index, ERDI) as presented in the ANSI/CTA-2034 spinorama plots, they are simply the listening window frequency response curve minus the sound power curve, or the early reflection curve.

Standard Method of Measurement for In-Home Loudspeakers (ANSI/CTA-2034-A R-2020)

This Standard describes how to determine the frequency response, directivity and maximum output capability of a residential loudspeaker.

shop.cta.tech

[Edit] For a more in depth discussion, see this article at Audioholics.

Understanding Loudspeaker Review Measurements Part I

Audioholics’ speaker reviews often contain detailed measurement graphs. In this article, we explain frequency response and the set of curves known as the 'Spin-O-Rama' to help you pick better product.

www.audioholics.com

 

[MattHooper]

Do you hear individual violins when you listen to an orchestra live (the assumption is it is a good orchestra). Answer: No, unless the conductor intends for you to hear an individual one (violin concerto), or there is one really bad violinist out of tune or tempo, or one that is audibly louder than the others. Let’s assume with our hypothetical orchestra, they perform like a well oiled machine . The concert master and other 9 violinists are in perfect unison and they sound as one and so audience members regardless of hearing level or skill can’t hear individual violinists.

Are we supposed to hear individual violinists in a recording of the orchestra?

Fortunately, that has a very easy and simple answer. It depends on how the orchestra was recorded. Then it depends on how the recording was mixed and what format - for sake of argument and simplicity, let’s say it end up being a 2 track stereo in digital format.

Exactly. Different recordings will have different sonic information.

Again, with professional grade microphones, proper placement, professional grade digital recorder, all of the information capable of being heard by the human ear (and much more) will be on that recording.

Will get back to that ^^^

Is a speaker capable of reproducing/transducing that content? If the speaker system has a relatively flat frequency response that covers the range of frequencies of the content (recording) then it is capable of reproducing (transiting it).

Ok, so it is your view that so long as a speaker displays flat frequency response, it can and will transduce any and all sonic information on a recording. In other words, there are no recordings whose sonic resolution outstrips a speaker that has flat frequency response. Is that right?

As I mentioned in the OP, I would expect at least some people to hold that view. Which makes a certain sense because, as I mentioned, Floyd Toole IIRC has said that things like audible speaker resonances will show up in frequency response. So if you have that flat, presumably it indicates a low influence of other factors like resonances. So if you think we already have speakers that measure flat enough, then there is no more left to do in terms of pushing various aspects of speaker performance (drivers/cabinets etc) in terms of getting all the detail from recordings.

Though there is obviously an interesting conundrum of "how do we know what detail is on a recording?" If the answer is "hearing it through speakers" that almost begs the question at hand: "how do we know the speaker is producing all the detail on the recording?"

Now, I don't actually think that is an insurmountable question at all. But I leave it for the moment to others in how to answer that question
(So, for instance, one could explain why we can know from a flat frequency response that a speaker is producing all the detail in a recording).

 

[Blumlein 88]

Though there is obviously an interesting conundrum of "how do we know what detail is on a recording?" If the answer is "hearing it through speakers" that almost begs the question at hand: "how do we know the speaker is producing all the detail on the recording?"

Now, I don't actually think that is an insurmountable question at all. But I leave it for the moment to others in how to answer that question
(So, for instance, one could explain why we can know from a flat frequency response that a speaker is producing all the detail in a recording).

I think headphones are a bit instructive. They can show detail, distortion, noise etc better than speakers (generally). Imaging is a weak point however. A lot of reasons for this, but part of it is you take the room out of the equation, usually any power issues, and probably more headphones are low distortion than speakers. Though I noted earlier that many good large speakers are in the .2 % THD range.

So good phones seem to provide more detail than good speakers. The room, room noise and cross talk between speakers seems likely to limit some detail retrieval that is on the recording to me. Seems likely though one cannot say it is proven. I also note it seems like good ESL speakers have more detail which I think is partly due to having a narrow dispersion and the dipole nature of panels takes maybe half the room out of the equation vs conventional speakers (plus they are low THD as well). Again, I would not call it proven, but seems to indicate the room as something of a limit in practice.

I suppose one could find a remote outdoor location quiet enough, and try good speakers without a room. You'll need to sit closer and have plenty of power, but does this get one better detail retrieval than conventional speakers in a room? It would be one test of how much the room is a problem.

I've noted in the past how modest speakers in rather large spaces seem to sound better and less colored than you expect. Modest speakers mounted in say a large open record store back when everywhere had record stores. Bars or strip clubs that had pretty decent sound if you had the chance to listen prior to opening when no one else was there. That sort of thing.

Supposition, conjecture, hypothesis......
If what I am thinking is true, do you engage something like dipoles, and put up with the deficiencies otherwise for more detail? Could other design criteria work around the room better? Can future DSP process out the room real time like the Klippel test rig processes it out for test signals. Maybe doing so with multiple channels most of which are just room cancellation channels. You have to figure out the room before you can find out if speakers are letting us down.

 

[FrankW]

Focusing on number of speakers is a separate issue from loudspeaker performance

Incorrect. It's critical to human perception https://www.aes.org/e-lib/browse.cfm?elib=14622

(accruacy)

Accurate to what perceptual reference?

I don't buy into the below-audibility blanket statements with the existing threshold studies.

Then provide supporting data

 

[FrankW]

Though there is obviously an interesting conundrum of "how do we know what detail is on a recording?" If the answer is "hearing it through speakers" that almost begs the question at hand: "how do we know the speaker is producing all the detail on the recording?"

Obviously?? 23+ pages and you finally start getting it
That circle is very confusing for some.

Bingo. Like "details" to be "retrieved". Audibility. What reference/absolute audible transduction "in the recording" are you comparing your transduced retrieval to for "details" "accuracy"? Never mind the complete illogic of stereophonic (loudspeaker transduction) "accuracy"

 

[FrankW]

I think headphones are a bit instructive.

Can be. But still transduction

 

[Blumlein 88]

Can be. But still transduction

Yes transducers are the problem. Transducers without the room vs transducers with the room. Plus single driver transducers that can cover the full spectrum.

 

[FrankW]

Yes transducers are the problem.

Could be. But the biggest problem is that there is no reference per se.
That and of course audiophile paranoia about trees while missing forest in a swirl of confusion.
There is no stereophonic loudspeaker "accuracy" again per se.

 

[Blumlein 88]

Could be. But the biggest problem is that there is no reference per se.
That and of course audiophile paranoia about trees while missing forest in a swirl of confusion.
There is no stereophonic loudspeaker "accuracy" again per se.

I already had in mind this is the sort problem best solved using monophonic transducers.

 

[IPunchCholla]

Can future DSP process out the room real time like the Klippel test rig processes it out for test signals.

Isn’t that kinda what BAACH does? Not the entire room, but at least the cross talk?

 

[MattHooper]

Obviously?? 23+ pages and you finally start getting it
That circle is very confusing for some.

This is hardly a new question. We've discussed the problem of accuracy from many angles, many times here and this was one of the implications that I had in mind starting the thread. When I asked "are we done yet?" and "where can we go from here?" obviously part of answering the question involves: How do we know when we've actually reproduced all the detail in a recording???

However, though the question certainly is relevant, as I indicated, it should be obvious that it is not a question that makes the concept of accuracy in speakers (in terms of reproducing recorded content) moot, much less insurmountable. We needn't be throwing up hands and saying "Ah well, I guess we can understand what it means to maintain an accurate signal from the perspective of A to D converting and D to A converting, and amplification all the way through the chain all the way up to speakers and then...gee who knows what could be accurate?" Do you really think we are in *that* position?

For instance, we have a good idea of how to maintain the signal accurately in our A to D conversion. And then in our D to A conversion. Then we have a good idea of what we want our amplifier to do, to maintain accuracy to the signal: Amplify the signal evenly across the relevant frequency spectrum, to maintain the balance encoded in the recorded signal. Well...that gives us a good idea of what we'd want to do when we get to speakers amplifying the signal: amplify the signal evenly across the relevant frequency spectrum. (As the Klippel paper says is the remit for good speaker design). This isn't some total mystery. If the skeptic wants to challenge this and say: "Well...how do we know even a perfectly neutral/flat speaker is accurately amplifying the recorded detail?" the question is turned back on the skeptic: Why WOULDN'T a speaker with flat amplitude/frequency response be accurately amplifying the signal? Especially given the logic of reproduction up to the point of the speaker?

As to why a speaker might not be accurately amplifying (via transducing) the information, well there we can appeal to various forms of distortion known...through testing...to obscure or change sonic information. Again...see the Klippel paper or countless other realms of information on this (and as an AES member, you should have some idea). And ask yourself the relevance of Toole suggesting standards from production to reproduction.

I'm not sure if we can end up with ultimate certainty we are hearing every detail possible in a recording. Perhaps there is an actual logical/physiological/technical answer to that question (which is one reason I've raised the subject to explore). But merely raising the question "how do we know we speakers are reproducing the recorded information?" doesn't actually keep us stuck in Plato's cave on the issue. It turns out we aren't actually stuck in purely circular reasoning. We can make practical/experimental inferences and predictions from a variety of angles - basically the stuff of audio engineering in the first place.

 

[mcdn]

Thank you so much. Do you happen to know what that directivity was, or name/citation of the paper(s) that was published in?

Any flat or shallow sloped index is good within reason, between the Schroeder frequency of the room and 20kHz. If it rises too steeply the radiated sound power drops too fast, making the speaker sound dark. There’s a lot of room for speaker designers to make choices though, based on the intended use and desired voicing.

 

[tuga]

Focusing on number of speakers is a separate issue from loudspeaker performance (accruacy), and one that is off topic in my view, same with discussing

Incorrect. It's critical to human perception https://www.aes.org/e-lib/browse.cfm?elib=14622

I don't have access to AES papers, is this another one of Harman's tests with a sample of 5?
Isn't it commonly accepted that single-speaker sound assessment of speaker performance is more discriminating than stereo?

accruacy

Accurate to what perceptual reference?

Accuracy to the recorded signal. It means the lowest possible levels of linear and non-linear distorions.
For example, speakers with non-coaxial tweeters have poor vertical directivity characteristics.

I don't buy into the below-audibility blanket statements with the existing threshold studies.

Then provide supporting data

I'd love to but I don't have the money to fund the research.

And you truly believe that audibility research is done and dusted?

 

[Travis]

Though there is obviously an interesting conundrum of "how do we know what detail is on a recording?" If the answer is "hearing it through speakers" that almost begs the question at hand: "how do we know the speaker is producing all the detail on the recording?"

I never said the way you can tell it’s in the recording is by listening through speakers, whether they are are flat, colored, or otherwise. It why I specifically used an example of 2 channel digital recording.

The information, all of it, is reflected graphically through software on a DAW. That information can be separated and isolated. My example included toe taps, baton tapping, HVAC hum, and a whole host of other things. Those are all real world examples, all of those thing can and have happened during a recording. It wouldn’t matter what speakers you played them through, nobody would be able to “hear” them in isolation because they are buried in the mix well below the general recording level of the performance. But all of it is there on the recording.

And how do we know that? We can take that 2 ch. digital recording and run it back through a DAW with something like Protools, run narrow high and low pass filters around the 21.7 hz and there is the HVAC hum, clear as day. The same with the toe taps, baton tapping, the defective hammer on the piano, (which might be identifiable by some with no isolation), and all the rest. The information/content is in the recording.

 

[tuga]

I never said the way you can tell it’s in the recording is by listening through speakers, whether they are are flat, colored, or otherwise. It why I specifically used an example of 2 channel digital recording.

The information, all of it, is reflected graphically through software on a DAW. That information can be separated and isolated. My example included toe taps, baton tapping, HVAC hum, and a whole host of other things. Those are all real world examples, all of those thing can and have happened during a recording. It wouldn’t matter what speakers you played them through, nobody would be able to “hear” them in isolation because they are buried in the mix well below the general recording level of the performance. But all of it is there on the recording.

And how do we know that? We can take that 2 ch. digital recording and run it back through a DAW with something like Protools, run narrow high and low pass filters around the 21.7 hz and there is the HVAC hum, clear as day. The same with the toe taps, baton tapping, the defective hammer on the piano, (which might be identifiable by some with no isolation), and all the rest. The information/content is in the recording.

I have a (non-audiophile) renaissance acapella choir recording where I can hear the birds inside the church. Once I heard them in a quiet part I could no longer not hear them unless the vocals were singing very loud an complex parts.
I'll try to find that recording.
I suppose close-mic'ed mixes will deal with those issues, but I generally don't like close-mic'ed classical.

 

[FrankW]

I'm not sure if we can end up with ultimate certainty we are hearing every detail possible in a recording.

We can't.
But if you want to answer a completely separate question of how to improve transduction via speakers, by improving the speakers themselves, that requires Post #13
Just like Toole, Klippel etc do and have done for decades.
That then begs the question of just how much improvement in speakers is needed, given what JJ was demonstrating decades ago...with "real" type auditory memory references, such as orchestra.
With 99% of studio fabricated music..

 

[FrankW]

I don't have access to AES papers, is this another one of Harman's tests with a sample of 5?

I know and yes, its another example of blind listening tests which you abhor.

Isn't it commonly accepted that single-speaker sound assessment of speaker performance is more discriminating than stereo?

Exactly. Thus corroborating my point about importance number of speakers when testing "Distortions". "Distortions" easily heard in one become more difficult with 2, close to zero by 5.

Accuracy to the recorded signal.

Which you have to transduce to hear thus have zero reference to be accurate to.

It means the lowest possible levels of linear and non-linear distorions.

No, because audible is all that matters.

For example, speakers with non-coaxial tweeters have poor vertical directivity characteristics.

Perhaps "poorer", just like our vertical hearing

I'd love to but I don't have the money to fund the research.

Poor excuse that never hinders your speculations

And you truly believe that audibility research is done and dusted?

Nope, but I believe dismissing 100 years of said research leads to looking audio forum audiophile silly, always without cognizance.

 

[tuga]

I know and yes, its another example of blind listening tests which you abhor.

I don't abhor blind tests. They are the only way to avoid sighted bias.

Like I said countless times, Toole got the blind test part right.
But it is obvious (to me at least) removing visual bias created other, unsurmountable problems which render the listening tests innefective or falsifies the results.

 

[FrankW]

But it is obvious (to me at least) removing visual bias created other, unsurmountable problems.

Yes, unwanted results.
Nice edit btw