I've been reading that amazing book by Floyd Toole (on psychoacoustics, "Sound Reproduction" 2008, 600 pp). In a couple of places he shows listener ratings of ESLs - particularly one very expensive speaker "M"..... Got that, "M"?
And they seem to come out with truly lousy approval ratings. Got to say, I like sizzling clean ESL sound and the other virtues and wouldn't think of other drivers since middle 70s.
So I wanted to start a thread, only partly psychotherapeutic, about why the ratings look so bad despite his methods that look so good.
Any ideas? Please no random attacks on blind listening tests but only if you're acquainted with his methods and see shortcomings in them.
And they seem to come out with truly lousy approval ratings. Got to say, I like sizzling clean ESL sound and the other virtues and wouldn't think of other drivers since middle 70s.
So I wanted to start a thread, only partly psychotherapeutic, about why the ratings look so bad despite his methods that look so good.
Any ideas? Please no random attacks on blind listening tests but only if you're acquainted with his methods and see shortcomings in them.
I was a little surprised by that myself. I don't listen to ESLs, but I do dipoles, almost exclusively.
AFAIK, this is the only time Toole compared a non-box speaker to other (box) speakers, and maybe just had a bad one. Or maybe there is some issue with the test protocol puts them at a disadvantage. Listener preference could also be at play (different types of polar patterns have inherently different presentations).
I really wish Toole would revisit non-box speakers - I, and a number of others, find them much preferable to box speakers.
AFAIK, this is the only time Toole compared a non-box speaker to other (box) speakers, and maybe just had a bad one. Or maybe there is some issue with the test protocol puts them at a disadvantage. Listener preference could also be at play (different types of polar patterns have inherently different presentations).
I really wish Toole would revisit non-box speakers - I, and a number of others, find them much preferable to box speakers.
Robh3606... if you are trying to say something, I wish you would just say it. I can't decipher your point, or even if you are making one.
To answer your questions straight, if that is what you intended (I can't tell), I have no idea what the other speakers are or if "M" stands for "Martin." Yes, I rather like Martin Logans. I've done a bit of blind ABX listening (in Stan Lipshitz' music room, no less) and certainly endorse the principle.
Toole's methods seem pretty clean. But here are some shortcomings I see immediately. First, he and Olive seem to find snippets from pop female singers to be the best test material. Besides the choice of genre, these are generally overblown fabrications from studio processing. Hard for me to imagine they would show off a great system well although he seems to think so.
True, some special material may result in numerically well-behaved stats. And Toole seems to dwell on that virtue in his tests and in his favorite testees, but that isn't the ultimate criterion of value for a test. Validity is the ultimate criterion of a measure.
Second, if you listen to 16 woofer-and-dome speakers, that sets your "adaptation level" just as sitting in bar with red lamps will make the white washroom garishly green looking. Many such illustrations in psychology textbooks. If then one speaker has a different nature (like ESLs), that will certainly sound odd.
Third, while Toole often says they have the same preferences, my sense is that his "trained" Harmon listeners may become "trained" to like certain sound qualities. He tries to say that certain physical measures duplicate the listener ratings but there is some circular reasoning in there and the correlation sometimes seems weak.
To answer your questions straight, if that is what you intended (I can't tell), I have no idea what the other speakers are or if "M" stands for "Martin." Yes, I rather like Martin Logans. I've done a bit of blind ABX listening (in Stan Lipshitz' music room, no less) and certainly endorse the principle.
Toole's methods seem pretty clean. But here are some shortcomings I see immediately. First, he and Olive seem to find snippets from pop female singers to be the best test material. Besides the choice of genre, these are generally overblown fabrications from studio processing. Hard for me to imagine they would show off a great system well although he seems to think so.
True, some special material may result in numerically well-behaved stats. And Toole seems to dwell on that virtue in his tests and in his favorite testees, but that isn't the ultimate criterion of value for a test. Validity is the ultimate criterion of a measure.
Second, if you listen to 16 woofer-and-dome speakers, that sets your "adaptation level" just as sitting in bar with red lamps will make the white washroom garishly green looking. Many such illustrations in psychology textbooks. If then one speaker has a different nature (like ESLs), that will certainly sound odd.
Third, while Toole often says they have the same preferences, my sense is that his "trained" Harmon listeners may become "trained" to like certain sound qualities. He tries to say that certain physical measures duplicate the listener ratings but there is some circular reasoning in there and the correlation sometimes seems weak.
Last edited:
The "M" speaker in Toole's test is the Quad ESL 63 if I recall correctly.
Although dispersion was the main design criteria for the Quad ESL 63 and it's probably better than most other ESLs, it's still not very good compared to modern 'constant directivity' speakers. It does have a distinctive 'sweet spot'. In his tests he has multiple listeners so inevitably not all of them can be in that sweet spot. In fact an ESL is probably the best example of a bad speaker according to Toole's criteria (if you don't mind my short summery): very flat frequency response, extended bass response and wide dispersion.
@bentoronto:
Toole also discusses the results of different test music and found that one guitar, one singer simple Jazz etc is not ideal for testing, but complex music like big classical orchestra's is. The latter showing defects better according to his findings. So I'm a bit surprised by your statement that he prefers 'snippets from pop female singers'. What is your source for this information?
Although dispersion was the main design criteria for the Quad ESL 63 and it's probably better than most other ESLs, it's still not very good compared to modern 'constant directivity' speakers. It does have a distinctive 'sweet spot'. In his tests he has multiple listeners so inevitably not all of them can be in that sweet spot. In fact an ESL is probably the best example of a bad speaker according to Toole's criteria (if you don't mind my short summery): very flat frequency response, extended bass response and wide dispersion.
@bentoronto:
Toole also discusses the results of different test music and found that one guitar, one singer simple Jazz etc is not ideal for testing, but complex music like big classical orchestra's is. The latter showing defects better according to his findings. So I'm a bit surprised by your statement that he prefers 'snippets from pop female singers'. What is your source for this information?
Yes, it was Martin Logan Vista, see here:
Audio Musings by Sean Olive: Some New Evidence That Generation Y May Prefer Accurate Sound Reproduction
Audio Musings by Sean Olive: Some New Evidence That Generation Y May Prefer Accurate Sound Reproduction
Please don't generalize. It is one particular model which gets one lousy rating. If you look at the corresponding respones curves, that rating seems to be well deserved.
Later on you talk about "adaptation level". If you really like the sound of the "M" speaker, you really need to think about your own adaptation level again
.His preference list does not look thaaat bad to me:
And everything is better than that ubiquitous jazz trio ...
I believe you have read this study too: www.aes.org/tmpFiles/elib/20100810/12206.pdf
As you can see in Olives "MP3 Preference Study" the Harman trained listeners are more discriminating than untrained students, but good stays good and bad stays bad in both groups.
If the test was done in stereo mode (which I don't know), all speakers were positioned 1.3 m in front of the front wall and 1 m from the next side wall. I have seen no information about toe-in.
Rudolf
Hello Sy
They are all in the same location during the testing. That could be an issue but we will never know if that was the issue.
Hello Bentoronto
I wanted to know if you have heard the speakers. Looking at the response graph they don't look all that great. I don't know why you or anyone else would think that the result of one speaker tested was an automatic condemnation of that speaker type. In this case ESL's.
They have done studies of trained vs untrained and the preferences are the same. The above posted link is one using college students.
Rob
I would never make a choice based upon Floyd Toole's work. But that is me.
I am not terribly happy with even the top of the line ML ESLs, but they are not terrible. Terrible you can get just fine through any decent ESL when you drive them with a signal chain that is "not quite up to the task."
Imo, a well configured ESL + signal chain ought to be noticeably easier to discern "differences" with than most speakers of other technologies... like 3-ways for example.
Ok, I did not read this study, but I have read quite a few of his over the years, and have been suitably underwhelmed... guess I'm not a fan?
_-_-bear
I am not terribly happy with even the top of the line ML ESLs, but they are not terrible. Terrible you can get just fine through any decent ESL when you drive them with a signal chain that is "not quite up to the task."
Imo, a well configured ESL + signal chain ought to be noticeably easier to discern "differences" with than most speakers of other technologies... like 3-ways for example.
Ok, I did not read this study, but I have read quite a few of his over the years, and have been suitably underwhelmed... guess I'm not a fan?
_-_-bear
No, but the test methods need to be understood before they can be interpreted. For example (going back to my question), since dipole ESLs are FAR more sensitive to positioning, orientation, and boundary reflectivity than conventional box speakers, trying to "equalize" the experiment by keeping speaker positions constant biases the experiment against the dipoles. Program material choice is also important- for example, the ESL57s I owned many moons ago were superb on string quartets but fell apart on "Dark Side of the Moon."
Hello Sy
OK fair enough. Could very well be placement was an issue. I am not going to buy into program material aspect though. A good speaker should not have issues with the types of music played.
Even if it was a case of "bad" placement that doesn't invalidate the results. All the speakers are listened too from the same room position. In that room position the others sounded better. That's a valid conclusion. If the position was optimized for a Dipole maybe the results would change.
Looking at the frequency response graphs that are on a slide in the presentation I don't think it was a placement only issue.
Rob
Last edited:
Well, the reality is that LOUD stuff is generally the Achilles heel of moderate-size ESLs. And ESL lovers are certainly quick to admit that. Give me a recording of a folk singer and a guitar, and a good ESL will sound marvelous. Add electric bass, electric guitar, and drums, screaming singer, synthesizer, and... well, not so much.
Gated measurements of frequency response can also be misleading. ESLs need just the right spot and just the right reflections- not only does their sound (especially spectral balance) rely greatly on the reverberant field, the thin diaphragms mean that reflections from the back wall pass right on through them. That's another big reason that the Procrustean methods which are valid for comparing one box speaker to another may not be applicable to speakers with a different radiation pattern.
To analogize, would the optimum spot for a cardioid, figure 8, and omni mike be in the same place?
100% agree.
Gated measurements of frequency response can also be misleading. ESLs need just the right spot and just the right reflections- not only does their sound (especially spectral balance) rely greatly on the reverberant field, the thin diaphragms mean that reflections from the back wall pass right on through them. That's another big reason that the Procrustean methods which are valid for comparing one box speaker to another may not be applicable to speakers with a different radiation pattern.
To analogize, would the optimum spot for a cardioid, figure 8, and omni mike be in the same place?
100% agree.
My 2-cents:
1. ESLs have their favored positioning but so do all speakers. Maybe ESLs suffered more in their "shuffle" mechanism locations. For sure, the narrow beam of ESLs (I sit pretty close to mine... you can do that with ESLs because they seem more transparent in the room, less in-your-face) prolly caused some off-angle listeners to be unhappy. The listening room was pretty large too. For sure, great bass seems to be the single more telling factor, Toole says.
2. From the price (many of Toole's speakers are megabuck), I guess that "M" as Martin Logan. But different Quads and other ESLs appear in the book and criticized. It would be nice for somebody to publish a key to identifying Toole's test speakers.
3. That chart of bestest program sources REALLY makes me suspicious. Can anybody can provide a rational for the sequence? As I mentioned before, their criterion for "good" is really "metrically good/stable" not "demonstrably valid".
For stat mavens re Olive's work: drawing conclusions from ANOVA and F-tests (or creating experiments where that's the outcome) is the kind of nonsense they preach in grad school; in Bus School the stat teachers say look at the means and SDs and what is the underlying process.
Some loose thoughts:
Gee, I wonder if modern cone and dome speakers are as bad as I remember them?
Looking at his loudspeaker chapter, makes it seem very hard to get the true flat compass and optimal directivity he seems to think is crucial; can anybody make speakers that good as DIY? Maybe ESLs, which need and get fairly little of the endless tinkering of his cone units, multiple crossovers, etc are a better route today instead of trying to achieve great sound with cones and endless fudging. I am blessed in owning full-range Dayton-Wrights - so I just tack on some treble and bass for an octave or two at each end with no crossovers in the large middle range.
About Geddes' mission-in-life to keep sound from bouncing off nearby walls: Toole takes lethal aim at Geddes' views about 80% of the time and supports them 20%... and somtimes seems to be doing so consciously.
Tastes good, burnt like Mom used to cook it: my jury is out on adaptation and personal-personal-personal preference and how they relate to "I like it" statements. And I have some graduate degrees in the subject.
1. ESLs have their favored positioning but so do all speakers. Maybe ESLs suffered more in their "shuffle" mechanism locations. For sure, the narrow beam of ESLs (I sit pretty close to mine... you can do that with ESLs because they seem more transparent in the room, less in-your-face) prolly caused some off-angle listeners to be unhappy. The listening room was pretty large too. For sure, great bass seems to be the single more telling factor, Toole says.
2. From the price (many of Toole's speakers are megabuck), I guess that "M" as Martin Logan. But different Quads and other ESLs appear in the book and criticized. It would be nice for somebody to publish a key to identifying Toole's test speakers.
3. That chart of bestest program sources REALLY makes me suspicious. Can anybody can provide a rational for the sequence? As I mentioned before, their criterion for "good" is really "metrically good/stable" not "demonstrably valid".
For stat mavens re Olive's work: drawing conclusions from ANOVA and F-tests (or creating experiments where that's the outcome) is the kind of nonsense they preach in grad school; in Bus School the stat teachers say look at the means and SDs and what is the underlying process.
Some loose thoughts:
Gee, I wonder if modern cone and dome speakers are as bad as I remember them?
Looking at his loudspeaker chapter, makes it seem very hard to get the true flat compass and optimal directivity he seems to think is crucial; can anybody make speakers that good as DIY? Maybe ESLs, which need and get fairly little of the endless tinkering of his cone units, multiple crossovers, etc are a better route today instead of trying to achieve great sound with cones and endless fudging. I am blessed in owning full-range Dayton-Wrights - so I just tack on some treble and bass for an octave or two at each end with no crossovers in the large middle range.
About Geddes' mission-in-life to keep sound from bouncing off nearby walls: Toole takes lethal aim at Geddes' views about 80% of the time and supports them 20%... and somtimes seems to be doing so consciously.
Tastes good, burnt like Mom used to cook it: my jury is out on adaptation and personal-personal-personal preference and how they relate to "I like it" statements. And I have some graduate degrees in the subject.
Last edited:
Hello Sy
OK fair enough. Could very well be placement was an issue. I am not going to buy into program material aspect though. A good speaker should not have issues with the types of music played.
Even if it was a case of "bad" placement that doesn't invalidate the results. All the speakers are listened too from the same room position. In that room position the others sounded better. That's a valid conclusion. If the position was optimized for a Dipole maybe the results would change.
Looking at the frequency response graphs that are on a slide in the presentation I don't think it was a placement only issue.
Rob
Rob,
the choice of program material is critical. I have a wonderful sounding CD that sounds WONDERFUL on just about anything it is played on or through. Actually it is something of a marvel in that regard. A horrid choice to discern "differences" if there ever was one. The source material matters greatly.
The big problem with these tests, imo, is that they ARE valid for the test, but NOT generalizable outside of the test conditions.
Frequency response graphs will tell you next to nothing about how a system sounds in an environment.
_-_-bear
Respectfully differ. What if the over-processed pop stuff has lotsa puff-up in the treble to give those female singers a sexy breathy sound (anybody think otherwise?). Sure would make speakers that are aenemic in the treble just right and ESLs that catch the last twitch on steel guitar strings (Ry Cooder's production of Alk Farkar Toure's "Talking Timbuctoo") sound ghastly.
I am painfully undecided: Toole and Olive have done a lot that right for an experiment and seem committed to "the truth." Very very hard to do these things beyond criticism. (Footnote about experiments: I am working this year with problems related to astronauts in space.... and no good way to simulate gravity-less work environments on Earth; likewise soldiers under deadly fire experiments).
Toole argues strongly that anechoic flat response is crucial (but that real-world bass loudness relates to liking more than anything else... too bad ESL fans). I don't know. For sure, no way to relate speaker design to a given music room below say 300 Hz.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.