A Broad Discussion of Speakers with Major Audio Luminaries

[MattHooper]

Thanks very much Dr Toole!

So if I infer correctly, on your view all or most speakers should converge to a certain sameness of sound with respect to tonal and timbral neutrality. Variations will happen within parameters like the size/frequency range of the loudspeaker, SPL requirements etc.

Ideally bass variations in rooms would be reduced via strategies such as well implemented subwoofers, perhaps room treatment, DSP etc.

To that degree, I suppose you would be in the “ yes generally speaking loudspeakers should sound the same” category.

With allowance made for individual preference in terms of manipulating EQ or tone controls to taste, or to correct for imbalances in recordings.

And also allowing for taste in terms of spatial qualities, in regards to how a loudspeaker interacts with the room (point source, wider dispersion narrower dispersion, dipole, omni)…. so long as those designs maintain low inherent coloration.

As for "omnis, dipole and bipoles, panel speakers, or any number of eccentric designs" these are all attempts to enhance the capabilities of stereo to address individual preferences. There is nothing distinctive in any of them that would make inherent resonances a desirable property - and our evidence from limited evaluations of these loudspeakers over the years is that they too need to be timbrally neutral, lending their "special" characteristics to the "soundstage and imaging" category of interrogation. In that category my investigations of many years ago - still not disproved - indicate that the recordings themselves are dominant factors.

(Bolding mine)

Having owned a wide variety of loudspeakers from narrow dispersion dipole/electrostats to dynamic speakers of varying dispersion all the way up to MBL omnis, in a 15’ x 13’ listening room…my subjective experience aligns with your conclusion.

Even when it came to the omnis, while they added a nice spacious “ floating free of the loudspeaker” impression, they did not fundamentally distort the character of the recording. Imaging relationships, the general shape of the sound stage remained pretty consistent when compared against my other speakers. And the spatial and reverb/acoustic cues in the recording were still dominant. There were significant and obvious differences between a tight, close mic’d “dry” recording that would put instruments “ in the room “ and a recording of instruments in a spacious hall.

 

[Salt]

They don’t read, or they read and misapply the research beyond what it was intended. It’s routine. In “their” defense, me included, there are interrelated but distinct concepts that are not always apparent that lead to this. One example: sighted listening bias; adaptation effect and need to use 3 or more pairs of speakers for DB testing.

So @Floyd Toole Dr. Toole, what should a consumer do to make the best decision on a purchase of floor standing speakers for their particular listening area?

Review Spinorama data
Make a short list
Get 2,3 or 4 pairs in to A/B blind?

4. Take home for test and listen

 

[Floyd Toole]

So @Floyd Toole Dr. Toole, what should a consumer do to make the best decision on a purchase of floor standing speakers for their particular listening area?

(a) read about the science - even that which has been explained in these forums. If it seems credible, go to (b), otherwise you are on your own.
(b) learn to interpret spinoramas or other comprehensive anechoic data. Some of the data one sees describes the loudspeaker output, not that which is most relevant to listeners - they are different things. Spinoramas focus on what the listener is likely to hear. Some of the colourful displays are good conversation pieces and are decorative, but really not very helpful in predicting what is heard in a room. Those that portray time domain information are sometimes misleading. In the absence of a properly set up blind listening comparison test I personally would trust a spinorama for guidance. Sadly, this kind of data is not available on all loudspeakers, but there are enough good ones where it is available that something can be found (do we really need all the loudspeaker brands that are in the marketplace?). From the perspective of sound quality the on-axis and listening window curves are very important - other curves should not exhibit evidence of resonances.
(c) part of (b) involves some speculation about the directivity and how it affects soundstage and imaging. Since both of these are strongly influenced by recordings, and some amount of personal preference is involved, there are no hard rules about directivity except that changes as a function of frequency should be gradual. Constant directivity is an unachievable goal with most practical loudspeakers as they all revert to omni at low frequencies. The most constant above crossover tend to be horns, but I know of no evidence that this characteristic is a benefit over well designed cone and dome systems.
(d) Set them up in your listening room, make low frequency measurements to identify room mode problems, and if possible fix them so that the virtues in the loudspeaker can be revealed. The methods are well described in my books and in Welti AES papers. Do not trust room curves above about 400 Hz, they are results, not targets.
(e) Enjoy. But be prepared to use tone controls or equalization for broadband compensation for recordings and/or room interactions.

In short, for consumers it is not easy. Sorry. Fortunately, there is abundant evidence that we humans can adapt to enjoy music through significantly flawed loudspeakers - until we hear something better.

 

[amirm]

Review Spinorama data
Make a short list
Get 2,3 or 4 pairs in to A/B blind?

It is beyond the scope of individuals to perform multi-pair blind tests.

I would trust the Spin and narrow the choice down further as far as looks, brand, cost, etc.

Then when you get it home, there is some possibility you won't like the tonality. If so, use EQ or tone control as Dr. Toole mentions to fine tune. You need EQ for bass anyway.

Net, net, an excellently measuring speaker that plays loud enough for you and is equalized in room has incredibly high chance of satisfying you.

 

[Floyd Toole]

Great audio engineers are not inconsistent in their work, as little as “trained listeners” are inconsistent in their work.

I would never disrespect the considerable talent of good recording engineers - some of them are brilliant! However, there is persuasive evidence that those among them with hearing loss are not consistent in their judgments of sound quality. A few have suffered to the point of having to change occupations. That is the dilemma these people confront - doing their jobs can result in a professional handicap.
Recording engineers who have graduated from colleges and university programs are trained on how to listen. But it is a very different kind of training from the "trained" listeners in our tests. They are trained to recognize audible resonances and to identify their frequencies to guide loudspeaker design engineers in attenuating them.

 

[Floyd Toole]

Some influential people, like the late Siegfried Linkwitz,

I knew Siegfried, and have listened with him in his living room, which was the space he designed his loudspeakers for and measured them in. He believed in science, and we had no fundamental disagreements. He did remarkable things without a financial sponsor and a lab. He is missed.

 

[amirm]

Where did I say anything about real vs creations?

Well, that is one standard of tonality being right. I was asking you if that is what you were using.

I don't need to know how close to the real thing they came in their creation, I can still judge the overall tonal balance of their creation. I find most audio productions to sound tonally well-balanced. If that wasn’t the case, there wouldn't be much point in me optimizing my sound system to a neutral response.

I don't find most recordings that way. There is a reason a tiny fraction of music out there is used at shows. I would say out of 30 to 50 tracks I listen to randomly in Roon, I add one to my reference list.

That aside, you need to actually work to determine if tonality is right. You need to change it to know if something else would have been better. You can't make an absolute determination as you just did. Even if you did, it wouldn't matter in this context. We need proof like this:

Here we see that the listener impression of tonality errors matched that of measurements (at lower resolution of course). These are the type of testers we can trust. Because we have tested them to know that they produce reliable results when we have the answers and know the right from wrong.

So please stop appealing to random authority. You need to put forward this type of evidence when challenged.

 

[MattHooper]

I would never disrespect the considerable talent of good recording engineers - some of them are brilliant! However, there is persuasive evidence that those among them with hearing loss are not consistent in their judgments of sound quality. A few have suffered to the point of having to change occupations. That is the dilemma these people confront - doing their jobs can result in a professional handicap.
Recording engineers who have graduated from colleges and university programs are trained on how to listen. But it is a very different kind of training from the "trained" listeners in our tests. They are trained to recognize audible resonances and to identify their frequencies to guide loudspeaker design engineers in attenuating them.

And on a similar note, obviously sound engineers and mixers are no less susceptible to bias than anybody else.

I’d say the majority of sound engineers and mixers I’ve known view audiophile obsessions with things like expensive cables as silly. However, I’ve also known some who are “audiophiles” and who buy in to “ cables sound different.” In fact there are even some boutique audiophile-oriented Studios outfitted with their “preferred” high end cables.

Personally, when it comes to all this stuff (listening without scientific controls) I like to try and keep two sides of the same coin in view:

The fact somebody may be perceptive in hearing certain real sound characteristics, doesn’t mean they can’t also imagine differences.

Also:

The fact that somebody can (or does) imagine certain sound characteristics doesn’t mean all their perceptions are inaccurate.

(So for example, even the engineers who imagine differences among expensive cables, may still be quite perceptive in their work, where it really counts).

(And of course good test protocols is what can allow you to sort through that to gain reliable knowledge).

 

[Petrushka]

I’m just going to repeat my opinion, that at a certain quality point, you can choose to listen to music rather than to defects.

With reasonably flat speakers, the defects will mostly be room effects or bass irregularities, which are likely to be room related.

I think that quality point can be reached at a reasonable price. Bass extension being the most expensive component.

 

[Salt]

I would never disrespect the considerable talent of good recording engineers - some of them are brilliant! However, there is persuasive evidence that those among them with hearing loss are not consistent in their judgments of sound quality. A few have suffered to the point of having to change occupations. That is the dilemma these people confront - doing their jobs can result in a professional handicap.
Recording engineers who have graduated from colleges and university programs are trained on how to listen. But it is a very different kind of training from the "trained" listeners in our tests. They are trained to recognize audible resonances and to identify their frequencies to guide loudspeaker design engineers in attenuating them.

A reasonable action to avoid 'hearing loss issues' would be yearly audiograms, not? (in some professions this is routine).

 

[sceptical1]

When evaluating speakers why not restrict the source material to live recordings instead of studio recordings where the COC is certainly apparent more often than not. Even so, I find this circle of confusion to be a matter of perpective, that is if the LP is changed with respect to the speakers any studio recording can sound much better than first thought. Simply, listening farfield when the recording was optimized in the nearfield must account for differing subjective opinions as much as the characteristics of any speaker.
What else matters most is the bass. If you want things to sound real +10 to +15db boost at 30hz in room at the LP is amazing. Much more effective as improvement than all the constant nitpicking about dispersion.
The output capability, quality, dynamics, and accuracy (implied, in terms of that personal preference, i.e. the harman curve) of low frequencies is all that distinguishes any speaker ever made from a google nest speaker.

 

[Floyd Toole]

(implied, in terms of that personal preference, i.e. the harman curve

Sorry, but I don't know what you are referring to when you say "harman curve" - there have been many over the years, all named by others, not Harman, and none of them have been related to any generalizable preference - personal, yes, and it varies. Yes, low bass extension is impressive and important, so long as it is not excessive. That was one of the first observations in my 1985/86 JAES papers.

When evaluating speakers why not restrict the source material to live recordings instead of studio recordings

Because classical recordings are not the most revealing of resonances and all recordings, classical/live or otherwise, get auditioned and frequently tweaked in control and/or mastering rooms before release. It is hard to avoid the COC.

 

[Newman]

[to Amir] You're the one who is expected to deliver adequate data here. For example, I have asked (or demanded) proper timing measurements quite a while ago, but nothing has really happened. First you didn't publish anything, then impulse response and nowadays step response, with occasional comments such as "I don't care about that" or "...fans of step response...". Neither can tell timing error 20-20k so I'm still hoping that you measure on-axis phase response and upload/distribute it for post-processing, and publish group delay, excess group delay and ETC. Please let us know if calculation of excess group delay is technically not supported or reliable enough due to missing features or inaccuracy of Klippel NFS. Other members such as I can fill missing features and produce the graph from uploaded magnitude and phase responses.

There is also some other missing data and/or evaluations. For example, recommended slope of power response (which produces recommended slope of directivity index), short and long term compression with spectrum closer to music (to be more comparable with practice). At the same time reviews contain graphs which are just informative. I don't mind that.

Of course it is possible to hide behind conclusions that effect is subtle and audibility is not guaranteed, but objective data should always be welcome due to few reasons:
- Perceivable effect and significance depend on type and magnitude of error. Without measurements we know nothing about possible error, and chances to capture explanation are lost for good.
- Available data opens possibility to make further investigations with statistics.
- Collecting objective data from distortions - even at low level - would show interest in development. Sticking into fixed bunker/corner with predefined filters does not.

Amir has responded with full rationality to such demands/requests as yours in the past. You should be able to find them or perhaps he will go to the unnecessary trouble of repeating himself.

One thing I will lead with: enormous reams of measurement data is exactly the wrong thing to do and very easy to do with modern measurement tools. Pruning it down to all the really important stuff is the job of the truly effective reviewer.

Perhaps you can lead with some backup of your demand/request, demonstrating (with evidence) the criticality of your requested additional data to choosing speakers that sound good.

cheers

 

[amirm]

You're the one who is expected to deliver adequate data here. For example, I have asked (or demanded) proper timing measurements quite a while ago, but nothing has really happened. First you didn't publish anything, then impulse response and nowadays step response, with occasional comments such as "I don't care about that" or "...fans of step response...". Neither can tell timing error 20-20k so I'm still hoping that you measure on-axis phase response and upload/distribute it for post-processing, and publish group delay, excess group delay and ETC. Please let us know if calculation of excess group delay is technically not supported or reliable enough due to missing features or inaccuracy of Klippel NFS. Other members such as I can fill missing features and produce the graph from uploaded magnitude and phase responses.

You can demand all you like that a Chinese restaurant give you a pizza, and a free one at that, but it won't happen. First, some of the things you are asking, Klippel NFS doesn't produce. Second, it is not my job to produce ton of data for individuals. It is to analyze performance of a speaker and allow myself and members to categorize it as terrible, maybe good, good, and great. You may have noticed some animal figures to that effect. Sure, I have gone further and sometimes performed analysis of a flaw but that is not part of the charter.

Similarly, I got tired of explaining why such measurements as waterfall and step response are useless in determining fidelity of speakers but eventually got tired and post them. They are adding to measurement time of a speaker which is already lengthiest of all the products I test.

Show me correlation to preference when it comes to those measurements with proper backing and not foot thumping and I will consider adding more. But don't look to me to donate more time to appease you and your demands. I have done my bit by investing so much money and time in this category, and giving away analysis data for people to use.

Finally, you are not the first person to think that the way to get me to add another test is by becoming belligerent. Maybe that has worked for you in the other aspects of your life but won't work with me. Learn to be persuasive by professionally explaining why something else would be good to test and I might just do that.

 

[BenB]

Because classical recordings are not the most revealing of resonances and all recordings, classical/live or otherwise, get auditioned and frequently tweaked in control and/or mastering rooms before release. It is hard to avoid the COC.

Didn't telarc avoid any tweaking of their recordings?

 

[kimmosto]

Klippel NFS doesn't produce.

Is it totally silent while moving the mic?

Second, it is not my job to produce ton of data for individuals. It is to analyze performance of a speaker and allow myself and members to categorize it as terrible, maybe good, good, and great.

What is your logic for example with near field measurements and impedance response? What is their audibility and significance at normal listening distances?

Similarly, I got tired of explaining why such measurements as waterfall and step response are useless in determining fidelity of speakers but eventually got tired and post them.

I totally understand. Those two are quite useless so you can leave them out and use the time for including phase to frequency responses or adding group delays.

Show me correlation to preference when it comes to those measurements with proper backing and not foot thumping and I will consider adding more.

What preference? At least timing was pre-filtered out from Olive's study and patent - if you mean that. The same with non-linear distortion, compression, magnitude of total directivity and slopes of power averages and directivity index. For example, recommendation for slope of power response was published few decades ago in some books (I have). Some say it was by Harman, but I haven't found the reference from AES eLib. There are many significant, but missing features in Olive's preference work. So "preference" argument does not look very strong imo.

Let me continue for a while with timing. I hope we could agree that (excess) group delay is audible above certain level. All studies I have found in AES library have come to this conclusion. Possible/probable problem is that most of listening tests are done with headphones so perceptions are limited to ears. Fortunately not all. So results are mostly not entirely valid for speaker reproduction due to missing tactile sensation which seems to be better sensor than ears especially at low mid and bass. Tactile sensation has threshold which compensates ear's weak ability to detect magnitude differences. Magnitude errors are obvious because timing errors corrupt magnitude of transients. Error could be radical in theory.
Key question is that do you want to trust that audibility is adequate reason to upload a few files or whole directivity data as 72 frequency responses including phase? No need to publish any more graphs. Personal opinions and experiences can be kept secrets though they seem to be as good as light science in this matter.

Finally, you are not the first person to think that the way to get me to add another test is by becoming belligerent.

Thanks for information. How about few last messages? Are they constructive enough or still attacking your lovely personality?

 

[tmtomh]

Thanks for information. How about few last messages? Are they constructive enough or still attacking your lovely personality?

What kind of response do you expect to get with that final sentence?

 

[kyuu]

What kind of response do you expect to get with that final sentence?

It's really bizarre when people think acting like an absolute tool is going to get them what they want.

 

[kimmosto]

At least I have pro measurement gear and I can capture requested extra information without problems so it's not about me and my needs. It's about getting wider objective, audible and possibly significant data at the same time when DUT is already available in front of mic. Without need to transfer the same products all around the world to some other who is motivated to export files with few different checkboxes checked. It's about attitude and will to serve community. Some could add also ecology and efficiency of the investments.

 

[Newman]

At least I have pro measurement gear and I can capture requested extra information without problems so it's not about me and my needs. It's about getting wider objective, audible and possibly significant data at the same time when DUT is already available in front of mic. Without need to transfer the same products all around the world to some other who is motivated to export files with few different checkboxes checked. It's about attitude and will to serve community. Some could add also ecology and efficiency of the investments.

Perhaps your greater error is in overestimating the importance of "requested extra information" in assessing the audible preference between loudspeakers. Start by demonstrating (with evidence) the criticality of your requested additional data to choosing speakers that sound good.