Originally Posted by Floyd Toole
Zzzzz... said: "Presumably the early prototypes had good spins. So I'm wondering why at least to some ears they didn't sound quite right while still measuring very well?"
Measurements are constants, stable and unchanging with time. Opinions are . . . different. I am not aware of this event, but it is quite possible that minor performance changes were involved . . . or not. I don't know what the man had been listening to and had adapted to. In my time on the job, I have seen experienced listeners, including audio professionals and reviewers, change their opinions dramatically based on "something" having nothing to do with the product, which was unchanged. In some cases, reviews especially, the change is attributed to "break in", or the like.
I almost hate to mention it, but it was professional recording engineers, engaged in double-blind listening tests evaluating studio monitors, who revealed the importance of hearing loss to me. It is discussed in a paper: Toole, F. E. (1985). “Subjective measurements of loudspeaker sound quality and listener preferences”, J. Audio Eng. Soc., 33. pp. 2-31. It is also explained in both of my books. Hearing loss is an occupational hazard in pro audio and it takes very little to cause easily observed and measured inconsistencies in subjective judgments. It is not uncommon for double-blind subjective ratings of the same loudspeaker to change dramatically over a few hours. It is not difficult to understand that if one is not able to hear low level timbral cues and distortions, opinions may be different from those of normal-hearing listeners, and may vary with time and program material. Critical listening is a task for young healthy ears :-( "Greybeards" should enjoy their music, but keep their opinions to themselves. I do - my hearing performance is shown in my books, and I am sitting here listening to my tinnitus which was partly sound and partly drug (OTC NSAID) induced.
Some people find it a problem that so much information is embedded in "simple" frequency responses. They persist in beliefs that something important is being ignored, usually related to non-linear distortion or waveform-related aspects, like phase shift, rise time, etc. Non-Linear distortions are important and much effort is expended to minimize them. In largish expensive loudspeakers it is rarely a problem. In inexpensive systems it is more common - good loudspeaker motors are expensive. The definitive evidence of a problem is subjective - distortion measurements commonly in use do not reliably correlate which what is audible in music. As I said, engineers aim for zero measured distortion, and get as close as possible within the cost, time and competence limitations.
Some discussions focus on a single factor, like diaphragm material, assuming that all audible differences are attributable to that. In fact, it is not uncommon for high priced tweeters using an exotic material, for example, to have superior motors yielding less non-linear distortion and reduced power compression. Differences may be heard that have nothing to do with the diaphragm - motors: the invisible variable. However, exotic materials have marketing attractions.
Waveform based factors are another popular topic of differentiation between loudspeakers. From an engineering perspective it is gratifying to see a clean impulse or step response, and to measure linear, or some other desired, phase response. From a listener's perspective it matters little or not at all. Humans are remarkably insensitive to phase shifts, many serious efforts to prove otherwise have not succeeded. Within drivers, which are minimum-phase devices, the time domain performance is predictable from the frequency response. There is no "speed" that is not explained by the frequency response and bandwidth at high frequencies.
Thanks so much for the comments. It seems that the spins are mostly all you need to judge the quality of a speaker. One part that has me a bit confused though is if the algorithm that determines how good a particular set of measurements is takes into account bass extension, why is there a discrepancy in correlations (0.8 vs 0.997)? Is it simply a result of not EQing the bass and differences in how each speaker's bass interacts with the room?
Regarding Frank Filipetti, I thought his opinion was interesting as he liked the earlier JBL LSR models, such as the 6328P, and I believe was still using them as his main studio monitors when he heard the prototype M2. (In fact he had done promotional videos for these monitors.) Also, it was Peter Chaikin that sought out his opinion on them. Recordings that Frank had worked on have been used in the Harman listening tests, and so.
You can read a bit about the early M2 prototype and Frank's opinion here:
One interesting fact is that the prototype didn't have the new waveguide and "beam[ed] high frequencies".
Edit: sorry I may have misrepresented what was stated in the above link. That the prototype "beam[ed] high frequencies" was not actually stated (he was talking about horns in general). He just states that he "didn't really like" the horn on the prototype.