A Broad Discussion of Speakers with Major Audio Luminaries

[Arindal]

the amount of boost should relate to the background noise in any specific vehicle.

Just out of curiosity: Did you have the chance to drive the i7, or current RR model based on the same platform? They are quiet to a degree you might want to forget about the necessity of masking any noise by bass boost.

when Canadian Broadcasting Corporation recording engineers participated in double-blind, equal loudness, multiple comparison evaluations of loudspeakers to choose new monitors their opinions agreed with those of audiophiles, and both groups preferred timbrally neutral, resonance- free loudspeakers. The best ones in the test at that time were consumer products, and the professionals were surprised by their choices, some demanding retests using their own master tapes - no difference.

I have taken part in similar (partly controlled) listening tests and even conducted them some 20 or 25 years later, with pretty similar results when it comes to preferred tonal balance. Surprisingly, designated studio monitors of this era were performing much better than high end or hi-fi speakers, which I would prescribe to their sound quality improving dramatically all the way since the early 1990s. Might have to add that all participants were either into recording of classical music, or from the public broadcasting sector, so particularly trained to judge timbre of natural instruments, with very little risk of having serious damage to their auditory system.

The preference for timbrally neutral speakers both on-axis and off-axis (translating to a constant directivity) was also pretty significant, with the exception of some recording engineers preferring the ´B&W sound´ at the time, with no difference between blind and sighted tests.

With respect to the B&W I wonder how your engineers were able to hear the absence of "edge diffraction problems"? Obviously they saw curved surfaces and assumed it.

B&W and other constructions aiming for the same absence of edge diffraction and narrowing directivity, were particularly lauded for:

- accurate localization
- stable localization
- excellent clarity/transparency/detail resolution in the brilliance bands
- wide ambience and excellent depth-of-field
- tonally balanced reverb

The two latter came to me as a surprise, admittingly, and might have been related to the pretty overdamped control room used. There was no difference to the results of the few blind tests, and no correlation between the baffle geometry and the praise for excellent localization/imaging. In fact, the other models which were pretty successful in this domain, looked quite edgy and boxy.

I'll bet nobody actually did anechoic measurements to confirm it. In fact the unbaffled tweeter is a problem, not a virtue. This is why most neutral loudspeakers use baffles and horns/waveguides on tweeters to match directivites at crossover.

We did anechoic measurements, even polar plots, to search for correlation, and yes, freely-mounted tweeters are always a source of inconsistent directivity, I agree. In overdamped rooms with the B&Ws (Matrix generation) of the time, it was not as dramatic as expected, though. The idea of matching midrange vs. tweeter directivity in the transitional band with the help of broad baffles or horns/waveguides is clear from a theoretical point. In practice, most of speakers following this ideal show an increasing directivity index towards higher frequencies, which in my perception leads to more severe problems than it is solving, particularly when judging recordings with natural reverb, as the additional reverb in the listening room inevitably becomes midrange/presence heavy and overly dull towards higher frequencies.

And yes, the B&W in question were delivering kind of the opposite, an overly brillant/bright reverb field (resulting from the tweeter´s pretty broad radiation in its lowest bands and a certain degree of suckout on octave below that), but this seemingly was not really bothering most of recording engineers, nor leading to tonal misjudgment of the recordings. Very few even liked it, and I mentioned them as award-winning engineers who are into classical music recordings and rank among the finest in the world.

If you have ever witnessed the live performance and broadcast downmix of a really really complex work in the control room (Mahler #8, some parts of ´Gurrelieder´ and ´Meistersinger´ come to mind), how a skilled mixing engineer is panning 40+ tracks from spot mics all over the place to a congruent and localizable sonic panorama just by ear, you might also come to the conclusion that these people know how to judge localization and stereo imaging of loudspeakers.

If any want to hear what their mix sounds like through an NS-10M, a B&W or anything else, which is not unreasonable, it can be accomplished by appropriately equalizing a neutral loudspeaker.

That might be the case with current B&W models more or less deviating from an ideal of linearity, but particularly the ones which were pretty well balanced on axis (like Matrix 801 Gen2), in my understanding were creating their specific sound mainly via directivity behavior and indirect sound. So I doubt this can be replicated via EQ, particularly when judging natural reverb on a recording.

The fact that the current trend is to close/near-field listening or listening in reflection reduced acoustically dead rooms makes the task easier - the direct sound dominates. The remaining problem is that consumers don't listen in the near field, so that cannot be the final evaluation.

The overdamped control rooms and use of nearfield monitors to judge a final mix in my understanding were mostly popular some years ago peaking in the 1990s. With control rooms getting additional surround channels being installed and more mixes focussing on imaging/ambience, I noticed kind of the opposite trend of more lively, bigger rooms in studios.

The problem with home listening conditions increasingly deviating from these ideals, of course remains, maybe getting even more dramatic with lots of tiny speakers, bigger and sparsely populated living rooms, increasing listening distances and reliance on digital room correction. Would have expected more of a movement towards effectively suppressing the listening room´s influence with the help of high-directivity index speakers, as they became very popular in sound reinforcement.

 

[Keith_W]

If a $1000 speaker measured like this (source: spinorama.org):

... I would probably find it in my heart to forgive. But the RRP is USD$23k. At that price, I expect perfection. There is no excuse! It's like going to a high end restaurant and the chef serves you well done steak! Don't these guys own microphones? Or have ears? I have heard these speakers, they are unlistenable. As the CEA2034 suggests, the only way to make them sound good is to have zero toe-in and try to absorb all that treble energy somehow. Otherwissssse they ssssound like a sssseveral sssibilant ssssnakessssss on ssssteroidssss.

 

[Sokel]

If a $1000 speaker measured like this (source: spinorama.org):

View attachment 502377

... I would probably find it in my heart to forgive. But the RRP is USD$23k. At that price, I expect perfection. There is no excuse! It's like going to a high end restaurant and the chef serves you well done steak! Don't these guys own microphones? Or have ears? I have heard these speakers, they are unlistenable. As the CEA2034 suggests, the only way to make them sound good is to have zero toe-in and try to absorb all that treble energy somehow. Otherwissssse they ssssound like a sssseveral sssibilant ssssnakessssss on ssssteroidssss.

My STIHL MS661 sounds way better (and gives me no round headache )

 

[audiofooled]

Award winning mixes have also been attributed to the use of NS-10s and Auratone 5Cs, which are not on anyone's list of loudspeakers to listen to for pleasure. Here is a quote from the Auratone website: "In 1982 Michael Jackson’s Thriller album, the best selling album of all time, was mixed on the 5C’s with Quincy Jones and Bruce Swedien. Bruce said that “80 percent of the mix was done on the 5C’s” and that Quincy calls them the “truth speakers."

I find many of the MJ's albums having track-to-track inconsistent tonal balance, being reproduced on a reasonably neutral system. It is hard to understand why until viewing measurements of such "truth speakers". Perhaps at the time lack of neutral reproduction systems was the driving force in accepting this "false truth", possibly granting an award.

But today? Why is use of such monitoring systems still being perpetuated as some kind of compatibility-translation check, side by side with neutral monitoring systems available?

 

[Salt]

My STIHL MS661 sounds way better (and gives me no round headache )

My personal preference is Husqarna, the harmonics sound more ... harmonic ...

 

[Floyd Toole]

Would have expected more of a movement towards effectively suppressing the listening room´s influence with the help of high-directivity index speakers, as they became very popular in sound reinforcement.

The problem is that all loudspeakers are omnidirectional at low frequencies, and any form of practical and uniform directivity control over the rest of the frequency range requires large horns. It has to do with wavelengths. Figure 12.9 from the 4th edition.

 

[Floyd Toole]

I find many of the MJ's albums having track-to-track inconsistent tonal balance, being reproduced on a reasonably neutral system. It is hard to understand why until viewing measurements of such "truth speakers". Perhaps at the time lack of neutral reproduction systems was the driving force in accepting this "false truth", possibly granting an award.

But today? Why is use of such monitoring systems still being perpetuated as some kind of compatibility-translation check, side by side with neutral monitoring systems available?

Check out Slide show 6 in the book website. The nonsense goes on, with new versions of these old speakers being promoted as somehow usefully revealing - of what? So much of this could have been ended long ago if people had any faith in the meaning of measurements instead of "from the hip" opinions. Trustworthy data on loudspeakers has existed but has not been widely available until recently, and it is getting better.

 

[MarkS]

You are misunderstanding or miss-characterising what people have stated.

There is one poster here who consistently and repeatedly states that sighted listening is completely and totally unreliable. That poster has not participated in this recent discussion.

I am happy to get confirmation from Dr. Toole that he disagrees with that claim. Dr. Toole wrote (concerning a sighted evaluation that I described) "But in this case, perhaps the difference in sound was great enough to overcome any residual bias."

It seems obvious to me that sound differences can be great enough to overcome any residual bias. I don't need a blind evalution to tell that a Revel Salon 2 sounds better than a Sonos Era 100.

And this is indeed confirmed not only by Dr. Toole's stated opinion, but also by data in his book, where the results of sighted evalutions clearly correlate with blind evaluations, but with significantly greater statistical uncertainty in the sighted evaluations.

 

[MattHooper]

The most trustworthy evaluations are those in which there is only one variable: the loudspeaker. The double-blind, equal-loudness, multiple (3 or 4) loudspeaker comparison method has been shown to be the most revealing and the most reliable.

The widely used "take it home and listen to it" method is the least reliable.

Regarding the sentence I have bolded:

Unless I’m missing something, for the consumer wouldn’t evaluating unfamiliar loudspeakers in an unfamiliar room (eg audio show, audio store, other audiophile’s home) be the least reliable method? Because you’re adding onto possible bias effects even more variables? At least with bringing a speaker home the room is a constant, as well as perhaps being able to check them against speakers you currently own, experience the different way the speaker will interact in your room, etc. (I know that visual biases can be strong enough to overcome even some room interaction… but surely that doesn’t rule out the possibility of hearing obvious new room issues with a new speaker).

Over the years like many audiophiles I’ve bought and tried a number of different loudspeakers in my room (for which there were no spinorama measurements, so this was pretty much the best method available to me). One pair of speakers that I liked when auditioning in a very large room at the store didn’t seem to be able to achieve the same sense of satisfying depth in my smaller room, and I could just never get over that so I sold them. Not to mention in direct comparisons with the speakers they were trying to replace, they couldn’t compete. This seemed to be informative in terms of whether those speakers were going to satisfy me in my own home or not. But… I certainly wouldn’t claim scientific certainty.

 

[Mort]

There is one poster here who consistently and repeatedly states that sighted listening is completely and totally unreliable. That poster has not participated in this recent discussion.

I am happy to get confirmation from Dr. Toole that he disagrees with that claim. Dr. Toole wrote (concerning a sighted evaluation that I described) "But in this case, perhaps the difference in sound was great enough to overcome any residual bias."

It seems obvious to me that sound differences can be great enough to overcome any residual bias. I don't need a blind evalution to tell that a Revel Salon 2 sounds better than a Sonos Era 100.

And this is indeed confirmed not only by Dr. Toole's stated opinion, but also by data in his book, where the results of sighted evalutions clearly correlate with blind evaluations, but with significantly greater statistical uncertainty in the sighted evaluations.

You've taken your victory lap. Won the big prize on the Internet. You were technically right!

 

[Newman]

There is one poster here

That's me. Other readers to note.

who consistently and repeatedly states that sighted listening is completely and totally unreliable. That poster has not participated in this recent discussion.

Incorrect. Misrepresentation. Lying BS. I am really sick of this. Here is my actual position. And written directly to you, too.

I am happy to get confirmation from Dr. Toole that he disagrees with that claim.

I will address that in my post below, to Dr Toole. Then he will see what you are really up to. It's unacceptable.

Dr. Toole wrote (concerning a sighted evaluation that I described) "But in this case, perhaps the difference in sound was great enough to overcome any residual bias."

It seems obvious to me that sound differences can be great enough to overcome any residual bias. I don't need a blind evalution to tell that a Revel Salon 2 sounds better than a Sonos Era 100.

I never said otherwise. I don't know why someone with your demonstrably high intelligence is willing to descend to this level.

And this is indeed confirmed not only by Dr. Toole's stated opinion, but also by data in his book, where the results of sighted evalutions clearly correlate with blind evaluations, but with significantly greater statistical uncertainty in the sighted evaluations.

...making them effectively useless in the home, ie untrustworthy. Which is the whole point that I have been making, and correctly quoting Dr Toole in support of such point, which you have been so determinedly misrepresenting.

So, you assert above that "sighted evaluations clearly correlate with blind evaluations"? Okay, show me any documentation of a correlation that can be used for general evaluation of loudspeakers, sighted. Who cares if there is a correlation in an experiment? Nobody said otherwise. That's bound to happen. It doesn't mean that there is a correlation that can be in any way useful for sighted evaluation of loudspeakers....unless you can show me one? Where has Dr Toole or any other noteworthy researcher documented a correlation that can be used for general evaluation of loudspeakers, sighted?

And if no such correlation is found, then what possible use do you want to make of the fact that sometimes a correlation exists, but unreliably so, ie we don't know exactly when that is going to happen?

Or is it because you want to work back to your indefensible positions like this, link? And this, link.

 

[Newman]

Dr. Toole, a poster here has claimed that you would agree with the following statement:

"Sighted listening is wholly and always unreliable.”

Do you agree with that statement?

And if you do, then can you please explain to me why I should prefer well-measuring speakers over poorly-measuring speakers. I listen to my speakers sighted. If I cannot reliably tell the difference between their sound under sighted conditions, why should I prefer well-measuring speakers?

Thank you.

When discussing subjective evaluations - opinions - one never speaks in absolutes. So I would like to think I never said that. The truth is that sighted listening opens the door to countless non-auditory influences on the judgement forming process. This is well discussed early in the 4th edition.

The most trustworthy evaluations are those in which there is only one variable: the loudspeaker. The double-blind, equal-loudness, multiple (3 or 4) loudspeaker comparison method has been shown to be the most revealing and the most reliable.

The widely used "take it home and listen to it" method is the least reliable.

You said: "If I cannot reliably tell the difference between their sound under sighted conditions, why should I prefer well-measuring speakers?" I cannot respond to this statement because we need to know the measurements on the loudspeakers you were comparing. If there is a large difference and you cannot hear it, perhaps you have a hearing problem. Differences can be present, sighted or blind, whether one responds to the acoustical evidence is partly dictated by non-auditory factors.

Hi Floyd,

Just to be clear, the person (MarkS) who asked you that question about sighted listening being “wholly and always unreliable”, has totally misrepresented what I said you said, link. And that's really sad. I can't help feeling a bit emotional about the way he has in the past consistently tried to misrepresent you and say that some of your experiments and your conclusions from your experiments are incorrect (see links below) and he is able to tell us the correct interpretation (specifically the one speaker vs stereo evaluation experiment and graph published in your first edition), and now he wants to misrepresent me as having told him that you said something that I never told him you said. I cannot help feeling that this is a sad state of affairs and we should not be forced to address this on ASR, but he is causing it to happen.

The fact is, here was the question that he actually asked, and it is not what he put in his question to you, that is only half of it. And in my answer, I provided answers which are quotes from you, that respond to the other half of the question. How sad is that?

I know how much you hate it when people twist and misrepresent your work and your opinions. Well, this man seems to be on somewhat of a mission. He has misrepresented your experimental data on sighted listening, link. For years he has insisted that the improved preference scores you published when listening to certain speakers in stereo is evidence that those speakers sound relatively better in stereo (compared to other speakers) than what they sound like in mono. And therefore in real practical use, the stereo result is the important one, link, link2. I have tried on multiple occasions to correct him and I have quoted some of your statements in this forum in support of that. This individual has a clear personal conclusion that stereo listening is as good as any for loudspeaker evaluation, and that listening generally at home to music in multichannel provides no important improvements to stereo listening, (also BTW that the multichannel recording catalogue is effectively zero), and he wants your views as stated throughout this thread, and in your books, about the improved acuity when evaluating loudspeakers as single speakers compared to stereo, to be wrong or at least unimportant. He has also stated on several occasions that his opinions cannot be wrong because they are drawing from his sighted listening experiences, and those cannot be so wrong as to mislead him. This is the fundamental issue here. He is making argumentum ad absurdum, by having you agree that the most extreme statement about sighted listening, i.e. that they never ever ever correlate with blind listening, is untrue. Well, of course it's untrue. And that's because nobody has claimed such, neither you, nor I, nor I in representing you. The fellow is throwing red herrings into the discussion and one has to wonder how he might use your response in future discussions about sighted listening impressions, eg partial quotes (again).

The real issue here, is that sighted listening cannot be trusted as a way to adjudicate sound waves themselves coming from loudspeakers. He wants that to not be true. He will IMHO probably be taking what you have said, that the most extreme overstated version of that fact is not true, to be at least somewhat extensible.

He wants to be able to walk away from this discussion thinking to himself (and probably asserting on the forums), “if my sighted listening impressions are fairly strong, then they are definitely representing what is in the sound waves themselves, and would correlate with a blind test.” Your gentle reply to such an assertion by him here, link, will be taken as encouragement.

I am hoping that such an impression is not one that you would endorse, nor intended to incidentally endorse with your response to the question when framed as “wholly and always unreliable”.

After all, you have made it clear in your books about your early days, how loudspeaker engineers were so confident that they could evaluate loudspeakers in sighted listening, with all their experience etc, and you put that notion to the sword. That is the enduring lesson that we general audiophiles need to take home.

You once said to me in a PM, and I don’t think I am betraying a confidence in repeating it here, “Anyone who thinks they can be unbiased in a sighted evaluation is misguided. That alone is unscientific.” This is the important takeaway, and any caveat along the lines that a sighted listening impression might sometimes also transfer to a blind test, is utterly inconsequential IMHO.

Cheers

 

[MattHooper]

...making them effectively useless in the home, ie untrustworthy.

I think you were doing pretty well up until this point. This is where you end up suggesting a more extreme level of scepticism.

“Less reliable” does not equate to
“effectively useless.” Yes if one is demanding scientific levels of confidence, in that respect they would be effectively useless.

Just like food scientists may find the informal cooking-at-home methods useless for the level of reliability they seek. That doesn’t therefore equate to the home cooking methods (“ maybe this needs more salt. I will try that”) are unreasonable or useless.

Nobody here - and I could be wrong, but I believe this includes MarkS - is arguing that sighted listening can provide the type of reliability you are demanding (or that science would demand).

The question is whether, in a practical sense, any useful information (which is going to have lower confidence levels than blind listening) can be gleaned from sighted listening.

For instance I’ve given a number of examples where I heard colorations under sighted conditions that told me I was unhappy with the speakers frequency balance, which were later corroborated in seeing the measurements. Had I seen the measurements first I may have written those speakers off without bothering to audition them. But even if I never saw the measurements my experience, it seems, had already correctly told me enough with respect to those speakers. And that doesn’t imply or require some perfect, always reliable impressions on my part, any more than it requires absolute certainty in the methods of a home cook.

Plenty of audiophiles have successfully used sighted listening in their home to identify some problematic areas, for instance, a troublesome bass emphasis, and used room measurements and room correction (or moving the speakers) to ameliorate those problems. Nobody is a perfect machine in this regard, but to say that nobody has gained useful information about their system through sighted listening would be extreme.
If that were the case…. Once again, we turned back to the problem of what would be the point of buying loudspeakers on how they actually sound?

I would also point again to Erin of Erin’s Audio Corner. If you pay attention to lots of his reviews, he appears to have quite a good track record of identifying various Sonic characteristics, whether it be neutrality or various colorations, in his sighted listening first which are then corroborated with his Klippel measurements afterwards.

Is this just a lot of luck? Or is it reasonable to infer that Erin has developed some level of skill in identifying speaker behavior, even under sighted conditions?

If somebody is seeking scientific levels of reliability, then it makes sense that they would not pay attention Erin’s subjective portions of the review at all, and just look at the measurements.

But given his apparent track record, if Erin reports that he tested a speaker and heard what he felt to be some obvious emphasis or scoop out in a certain frequency, I don’t think unreasonable to, informally speaking, think
“there’s a good chance he’s on something there.” Even if there were no measurements to look at.

I certainly could be wrong about any of the above. But that is at least represents the view, generally being put forth I believe.

 

[kyuu]

If a $1000 speaker measured like this (source: spinorama.org):

View attachment 502377

... I would probably find it in my heart to forgive.

You're way more tolerant than I am. I wouldn't accept that in a $500 speaker (and I didnt when I purchased mine), let alone $1k.

 

[Holmz]

There is one poster here who consistently and repeatedly states that sighted listening is completely and totally unreliable. That poster has not participated in this recent discussion.

I am happy to get confirmation from Dr. Toole that he disagrees with that claim. Dr. Toole wrote (concerning a sighted evaluation that I described) "But in this case, perhaps the difference in sound was great enough to overcome any residual bias."
…

Seriously?

There are books and talks going back 40 years, why do you need to confirm this now?
Are you suspecting some late onset shift?

 

[Jiraya369]

If a $1000 speaker measured like this (source: spinorama.org):

View attachment 502377

... I would probably find it in my heart to forgive. But the RRP is USD$23k. At that price, I expect perfection. There is no excuse! It's like going to a high end restaurant and the chef serves you well done steak! Don't these guys own microphones? Or have ears? I have heard these speakers, they are unlistenable. As the CEA2034 suggests, the only way to make them sound good is to have zero toe-in and try to absorb all that treble energy somehow. Otherwissssse they ssssound like a sssseveral sssibilant ssssnakessssss on ssssteroidssss.

well it's not like the salon 2s are flat either at that price but that doesn't stop it from being perceived as great if not SOTA and some calling it "perfect". I think at some point the measurements need not be absolutely perfect and just reasonably linear, other things like loudness with respect to distortion/compression and more importantly low end extension ought to aid significantly in perceived quality/preferences. Idk maybe im just yapping and i dont care too much about perfect tonality. +-3 db that is reasonably linear is good enough to me, but maybe i've just listened to good speakers lol

i agree though the response presented is horrid for any sane person but the directivity is reasonable, a simple eq job ought to help it for the most part, but still it goes up and down way too hard and one shouldnt have to at this price point. The thing is, this bright response IS what sells. It's "detailed", "revealing" and all other buzzwords ignorant people can think of. I once heard someone call a B&W 800 series speaker "too neutral" and "too revealing". You can see the problem here, they believe one thing and think this boosted treble is neutrality somehow lmao

 

[Arindal]

The problem is that all loudspeakers are omnidirectional at low frequencies, and any form of practical and uniform directivity control over the rest of the frequency range requires large horns. It has to do with wavelengths.

Both problems have in my understanding been largely solved over the course of the last 15 years, in studio monitoring, sound reinforcement and high end audio alike, thanks to fullrange cardioids and line sources. Yes, the max SPL of the former is somehow limited, and the effectivity of the latter depends on the relative length of the line to the wavelength, but this is a common thing in ultra compact P.A. systems. Examples:

 

[Arindal]

it's not like the salon 2s are flat either at that price but that doesn't stop it from being perceived as great if not SOTA and some calling it "perfect".

By no means supporting that position, but I have met a significant number of people who name the B&Ws SOTA and perfect.

I think at some point the measurements need not be absolutely perfect and just reasonably linear,

Absolutely agree to this point, particularly when we are talking about narrow-banded deviations which can look pretty ugly on the graph but sometimes turn out to be negligible. When discussing solely based on measurements, which speakers meet the aims formulated by Dr. Toole the most, I would start with looking at smoothed graphs and weighting broad-banded deviations stronger, both under anechoic conditions and under different angles.

+-3 db that is reasonably linear is good enough to me, but maybe i've just listened to good speakers

If you have an octave band at -3dB and the neighboring bands +3dB, that might a still meet the +-3dB tolerance band from mathematical point, but I would expect such speaker to sound severely colorated. Particularly if we are talking about the bands in which our ears are most sensitive and our brain can differentiate direct sound from reflections to a certain degree.

The thing is, this bright response IS what sells. It's "detailed", "revealing" and all other buzzwords ignorant people can think of.

This might have been the case with high end speakers in the 1990s demo´ed with minimalistic acoustical recordings, folk and jazz and chamber music, enjoyed by many audiophiles who were on the quest for ´maximum of detail resolution´.

I have my doubts if this is still the case. Particularly modern rock and pop music has gotten pretty treble- and brilliance rich over the course of the last 35 years, and of course mastering engineers, radio guys and producers also came to the conclusion that enhanced treble/brilliance sells. Playing such recordings over bright speakers of any kind bears a pretty high risk of at least a number of tracks sounding overly bright, fatiguing, thin and sibilant-heavy. If we take into account, that listening to internet radio, algorithm-based streams like Spotify radio, or playlists compiled by someone else, are increasingly popular, I would expect people to rather dislike overly bright speakers on the long run.

While some manufacturers seems to tend towards brighter-sounding speakers, others seem to reduce treble and brillance in the room deliberately to an astonishing degree (which I personally find way too dull and equally annoying). If there is a general trend recently among consumer-class speakers, seemingly a lot aim for deeper, fatter and more voluminous bass. If I would have to sum it up I would say that we have several contradicting trends at play, ironically leading to current speakers sounding increasingly different from each other, although technical progress might suggest they could easily all come closer to one ideal.

 

[rdenney]

It seems to me that there is a range of performance measures for speakers that affect their ability to meet needs in any particular use case. In no particular order:

1. Lack of resonances. We don't want speakers ringing at certain frequencies for whatever reason--such is often audible and annoying.

2. Reasonably flat anechoic on-axis frequency response, even if it requires equalization. If the on-axis response isn't flat, and we make it even more non-flat to compensate for some non-flat influence from the room, we end up with flat frequency response from only one reliable location. We might get lucky at other locations.

3. Directivity that is tonally sensible off-axis. That means that the tonality of what is sent off-axis bears close enough relationship to what is presented axially that the reflections in the room seem realistic.

4. Dynamic contrast. If the compliance around the drivers becomes elastically non-linear when reaching the limits of its excursion (and assuming the voice coil doesn't mechanically bottom out), the frequencies provided by that driver will not be as loud relative to lower levels as is the signal from the recording. This changes the spectral response if it only happens to one driver, and it will probably affect lower frequencies more than higher frequencies. If all the drivers behave similarly (they won't), then the speaker might simply be unable to play the loud bits as loud relative to the soft bits as was provided on the recording, but without an obvious tonal change. The bigger the room and the higher the demand for listening level (and the smaller the speaker), the more likely this effect will become a noticeable fault. Just like amp clipping, this seems to me to overwhelm other effects in those cases where it is an issue. Isn't this what the additional bass driver in the Salon towers provide compared to, say, the Concerta towers--the ability to get louder without the the bass drivers running out of steam? (Maybe a bad example: Soundstage's measurements at the NRC anechoic chamber of the two didn't note much difference in bass distortion, and the F12 actually had better linearity with respect to 70 dB output when playing at 95 dB. About 5 dB greater sensitivity, too. The Salon also has an additional mid-range driver and much lower distortion around 1KHz, however--that's the significant difference.) Isn't this one of the things that distinguishes large speakers from small speakers? I suspect this effect was not measured in most of Floyd's work, but I may have just missed it. I doubt his preference tests were conducted loudly enough to test compression, especially in the smaller test room. I don't think subwoofers, which are often crossed over well below the crossover between the woofer and the higher drivers, can make up for this if the speakers show a lack. Many will not undertake this use case, but some certainly do. I really don't think my Concerta F12's could make the same dynamics as, say, an Altec A7, which was designed for a much different use case with much larger rooms, even though I know I can play them very loudly.

5. Lack of distortion. At some point, distortion becomes audible. If the distortion in the bass driver is at, say, over 10% (not really that uncommon), the distortion components will be hearable in a comparison, sighted or not, it seems to me. I can hear it. I think this is what makes tubas and french horns sound like trombones--an effect I have experienced in a listening test when I had no knowledge of or interest in the different speakers being demonstrated. The addition of high-frequency distortion products can make that tonal difference.

6. I hesitate to add this, because I've never detected any difference between speakers I could attribute to it: time alignment.

Some of these things are definitely audible. I've heard speakers that were so emphasized in high frequencies that they had a frying-bacon effect. I've heard speakers that distorted so easily in the bass that tubas became trombones or euphoniums when the listening level was increased (NOT when the performer was playing louder). I've heard speakers that muffle loud transients because the driver excursion limits were being reached. I've heard systems that went from realistic to crazy just by walking across the room, even in a room where live performers sound much more consistent from different listening locations. In the times I have set up sound systems, such as at my church, I used careful equalization using calibrated microphones and loudspeakers selected with considerable intentionality, and still noted that in this back corner, treble was suppressed, etc. Having the measurement capability didn't prevent me from using my ears, though it did prevent me from trusting my ears too much.

Finally, about preference testing: The preference results were not unanimous, and the correlations were reasonably strong from a statistical perspective only in the realm of human sensitivities and opinions, where unexplained variability is the frequent result. They describe a sampling of preferences that, because of those statistics, are demonstrated not to be the product of mere chance but that are actually representative, at some level, of the population at large. If we describe those preferences as a random variable described by some probability distribution, we can use the results as a model of what most people prefer. Of course, all models are false, even if some are useful. But it does not state what I may prefer, or what you may prefer, if we, for whatever reason, find ourselves on the tails of that distribution. As such, it's much more instructive to manufacturers because it tells them what the center of the market prefers and therefore what they should design for. It's not necessarily instructive to me, individually, except to suggest that I may have trained myself, intentionally or not, to have preferences outside the norm and perhaps I should consider that.

So, when Floyd makes statements like "people like bass," he's describing a reasonably well-correlated effect from studies, but not necessarily Rick. "People" may like a lot more bass than, say, tuba players, who live in the bass world and want to hear distinctions that may be lost on others. "People" may like bass even when it is distorted at 10% or more, even though the instrument has characteristic overtones at the same levels (20 dB down) as the loudest distortion products. Example: Mr. Gene Pokorny (of the Chicago Symphony) recorded a CD full of orchestral excerpts some years ago, for the benefit of tuba players. In some of the excerpts, he used a B&S F tuba, and in some, he used the CSO's York C tuba that was previously owned by Arnold Jacobs. Most people can't tell the difference listening to the recording, despite that the two instruments are vastly different in size and breadth of sound. Any tuba player can. I've heard well-recorded Youtube videos of orchestral performers comparing similar grand orchestral tubas to that York, and heard distinctions in headphones that speakers in a room did not reveal. Most people would neither care nor notice, of course. (I am certainly NOT suggesting that musicians make good audio system listeners--in my experience their systems don't consistently sound good and it may be that musicians are just good at filling in the blanks from their experience. I'm just suggesting that musicians may be very specific in what they are listening for that is not represented at all in the sorts of preference testing done by Toole, Olive, and so on.)

Ears do matter. The problem is that they are so dominated by eyes and brains, which are easily and uncontrollably swayed by inaudible effects. But that we can't fully trust our ears for making choices in the presence of sighted bias doesn't mean that we still don't use our ears to hear and make judgments about what our systems produce.

Rick "has said all this before" Denney

 

[Floyd Toole]

Hi Floyd,

Just to be clear, the person (MarkS) who asked you that question about sighted listening being “wholly and always unreliable”, has totally misrepresented what I said you said, link. And that's really sad. I can't help feeling a bit emotional about the way he has in the past consistently tried to misrepresent you and say that some of your experiments and your conclusions from your experiments are incorrect (see links below) and he is able to tell us the correct interpretation (specifically the one speaker vs stereo evaluation experiment and graph published in your first edition), and now he wants to misrepresent me as having told him that you said something that I never told him you said. I cannot help feeling that this is a sad state of affairs and we should not be forced to address this on ASR, but he is causing it to happen.

The fact is, here was the question that he actually asked, and it is not what he put in his question to you, that is only half of it. And in my answer, I provided answers which are quotes from you, that respond to the other half of the question. How sad is that?

I know how much you hate it when people twist and misrepresent your work and your opinions. Well, this man seems to be on somewhat of a mission. He has misrepresented your experimental data on sighted listening, link. For years he has insisted that the improved preference scores you published when listening to certain speakers in stereo is evidence that those speakers sound relatively better in stereo (compared to other speakers) than what they sound like in mono. And therefore in real practical use, the stereo result is the important one, link, link2. I have tried on multiple occasions to correct him and I have quoted some of your statements in this forum in support of that. This individual has a clear personal conclusion that stereo listening is as good as any for loudspeaker evaluation, and that listening generally at home to music in multichannel provides no important improvements to stereo listening, (also BTW that the multichannel recording catalogue is effectively zero), and he wants your views as stated throughout this thread, and in your books, about the improved acuity when evaluating loudspeakers as single speakers compared to stereo, to be wrong or at least unimportant. He has also stated on several occasions that his opinions cannot be wrong because they are drawing from his sighted listening experiences, and those cannot be so wrong as to mislead him. This is the fundamental issue here. He is making argumentum ad absurdum, by having you agree that the most extreme statement about sighted listening, i.e. that they never ever ever correlate with blind listening, is untrue. Well, of course it's untrue. And that's because nobody has claimed such, neither you, nor I, nor I in representing you. The fellow is throwing red herrings into the discussion and one has to wonder how he might use your response in future discussions about sighted listening impressions, eg partial quotes (again).

The real issue here, is that sighted listening cannot be trusted as a way to adjudicate sound waves themselves coming from loudspeakers. He wants that to not be true. He will IMHO probably be taking what you have said, that the most extreme overstated version of that fact is not true, to be at least somewhat extensible.

He wants to be able to walk away from this discussion thinking to himself (and probably asserting on the forums), “if my sighted listening impressions are fairly strong, then they are definitely representing what is in the sound waves themselves, and would correlate with a blind test.” Your gentle reply to such an assertion by him here, link, will be taken as encouragement.

I am hoping that such an impression is not one that you would endorse, nor intended to incidentally endorse with your response to the question when framed as “wholly and always unreliable”.

After all, you have made it clear in your books about your early days, how loudspeaker engineers were so confident that they could evaluate loudspeakers in sighted listening, with all their experience etc, and you put that notion to the sword. That is the enduring lesson that we general audiophiles need to take home.

You once said to me in a PM, and I don’t think I am betraying a confidence in repeating it here, “Anyone who thinks they can be unbiased in a sighted evaluation is misguided. That alone is unscientific.” This is the important takeaway, and any caveat along the lines that a sighted listening impression might sometimes also transfer to a blind test, is utterly inconsequential IMHO.

Cheers

I thought all this was past, but I wasn't aware of the prehistory. My understanding of recent events can be summarized as follows:

MarkS bought a pair of AsciiLab C6B because their anechoic measurements were more attractive than those of his Golden Ear Tritons. He listened to them sighted and preferred the AsciiLabs.
He entered the sighted listening test with a bias, having spent money on what was likely to be a better sounding loudspeaker, and when he listened he heard a better sounding loudspeaker.
All that is at issue is how much of his preference is related to the sounds arriving at his ears, and to what extent that opinion was biased by his prior knowledge. We will never know.
The reality is that he selected his preferred loudspeaker on the basis of anechoic measurements, to which I say "bravo"!

If I have a final word on the topic let it be: The factors influencing opinions formed in all listening tests involve acoustical and non-acoustical factors. Controlling the auditory factors is difficult. Comparisons require equal loudness, rapid changes, and positional substitutions if room effects are to be normalized. When this is done in double-blind fashion there are decades of results showing usefully reliable opinions from hundreds of listeners of all backgrounds. The principal factor causing statistical variations has been hearing loss.

Sighted evaluations introduce the possibility that opinions will be biased by non-auditory factors. The extent of the bias depends on the nature and importance of the non-auditory factors to the individual listeners.

In science one does all that is possible to eliminate biasing factors. End of story. Cheers!