Thanks for posting Carp. One of the more interesting posts below. I don't ever remember Dr. Floyd Toole every commenting on my ability to conduct a GTG
A "Toole" summing up
Hello again. I have waited until you had a chance to vent your feelings before commenting. John has also asked a couple of specific questions of me that I think will be interesting to some of you at least.
First let me say that I am greatly reassured by the result: two truly excellent loudspeakers were judged to sound as good as their measurements suggest they should. In the global context of loudspeakers both are in the highest category. I regard it as a minor breakthrough that; (a) measurements were acknowledged as having meaning in terms of sound quality, and (b) doing listening tests blind was recognized as allowing one to form opinions free from non-auditory influences. None of this is new, but only a tiny minority in the audio community have bothered to read at least some of the existing science. That includes many of this group – and thanks!
What John and his colleagues created, and what several of you participated in, was a reasonably well controlled scientific experiment. It took knowledge, discipline, preparation and hard work – so much that it is rarely done. It is why run of the mill “subjective” reviews are so poorly done, and the results so untrustworthy. It woud be nice to have more of this. More importantly, more people need to be made aware that there are measurements, the Spinorama being perhaps the best one, that do a remarkable job of describing how a loudspeaker might sound. Sadly, these measurements are scarce. Many manufacturers can’t do them, and most would not like what they saw if they did them. Others seem not to have a conscience, relying on marketing and folklore, including the myth that “we can’t measure what we can hear”.
This is why I wrote the first book, and now its successor. I am currently creating content for a companion website, where page count doesn’t matter. There is much to read that I think will answer your questions, and more, including some of my earlier posts in this forum. But, for now, here are some quick comments.
Mono vs. Stereo. I commented on the mono/stereo issue earlier in this thread, and it is discussed in both books. Stereo is more entertaining than mono, but mono is more revealing of problems in loudspeakers. A good loudspeaker in mono is also a good loudspeaker in stereo. Imaging is dominated by recordings, which are control-room creations – yes, even classical ones – judged through loudspeakers.
Why doesn’t it sound “real’? Because it can’t. A small number of microphones cannot capture the 3D sound field around performers, and two channels and two loudspeakers of any design cannot reconstruct a realistic sound field. It is the daunting task of recording engineers to do what is possible to deliver a semblance of something real, if that is the goal, or to create an artificial “stereo” variation that is stimulating and entertaining. We need multichannel.
Hearing performance. Many years ago, I ran a test for the Canadian Broadcasting Corporation. The purpose was to select small, medium and large monitor loudspeakers for use throughout the national network. They should be good and sound alike – if possible. The listeners were recording engineers and producers from across the nation. The results are in JAES papers and the books. Several important lessons were learned by all. The sad one was that several of the recording engineers had significant hearing loss and this was exhibited as large variations in their opinions during the double-blind listening tests. Some bias was also seen. Why? They weren’t hearing all of the small details in the music, nor all of the small defects in the loudspeakers. Lesson: use listeners with hearing thresholds within 20 dB of normal – which includes about 75% of the population, a sizeable target audience. If a person has hearing loss, those opinions are for that person alone. But what if that person is a reviewer . . . ?
Trained vs. untrained listeners. The most common and most audible problem in loudspeakers is resonances. Most people hear them, but few can describe them. Sean Olive created a training program to teach people what resonances of different Q, frequency and amplitude sounded like and tested their ability to hear and report on them. Because the results of the listening tests were fed back to the engineers, it was useful that these trained listeners could describe what they heard in terms that were useful to the design engineers. There was a resulting bias from this training: it was that these listeners were less easily impressed than average customers. The consequence has been that they award lower sound quality ratings, but the order of their ratings is the same as those from hundreds of untrained listeners that have been tested.
Room interactions. At frequencies below about 300-500 Hz in small rooms the room dominates the quality of bass, and bass accounts for about 30% of the factor weighting in subjective evaluations of sound quality. That is why we do listening tests using positional substitution, so the room is a constant factor. Multiple subwoofers, signal processing and judicious equalization can greatly improve things, to the point that the room “disappears” – there are no audible resonances.
Above about 500 Hz the loudspeaker itself assumes the dominant role. The explanation is complex, but it comes down to the fact that humans are well adapted to listening in reflective spaces. To a substantial extent it is possible to “listen through” a room. In live, unamplified performances, we hear familiar voices and instruments in different acoustical settings. They remain the same voices and instruments, but in different venues. In sound reproduction, the parallel situation holds, but the loudspeaker is a proxy for the voices and instruments. The success of the reproduction is totally dependent on the skill of the recording engineer in capturing and storing the timbral essence of the live sources, and on the skill of the loudspeaker designer to create a “neutral” transduction device so that we can hear what was recorded.
This is the circle of confusion wherein professionals do not always use neutral monitors and consumers don’t always have neutral reproducers. In the experiment that John just conducted, we had a professional monitor loudspeaker and an audiophile loudspeaker. If it is judged that they sounded similar, then Harman has contributed to eliminating the circle of confusion. Recordings mixed and/or mastered using either of these loudspeakers would sound quite similar on the other.
In order for this “listening through the room” to happen, one must have some familiarity with the room - more is better - but it has been shown that a few minutes are sufficient to show improved speech intelligibility and the like. Early reflections are the dominant room interactions above 500 Hz and this may be the most significant difference between these loudspeakers – directivity. They are quite close, but the Salon2 will energize more far-off-axis early reflections than the M2. For laying down tracks and mixing many engineers discourage room reflections, even use “near field” monitors, although mastering engineers often prefer to hear some “room”. I discuss this in some detail in the book, and it is clear that “one size does not fit all”. Some evidence points to hearing loss (an occupational hazard in pro audio) as one factor causing professionals to prefer simple, reflection-free, sound fields – again details in the book.
Would changing rooms have changed the results? Probably not in any substantial way – even for these closely rated loudspeakers. In evaluations of loudspeakers in general, moving the evaluation from room to room the numbers may vary, but the order of preference remains. If you look at the geometry of a listening situation it will be found that the early lateral reflections are generated by sounds radiated at large angles off axis – moving back in a room varies the direct-to-reflected ratio, as more sound from smaller angles contributes.
Room EQ. When we measure in a room we use an omnidirectional microphone that ignores the angle of incidence of incoming sound, and, if one is doing steady-state measurements, also the timing. Two ears and a brain are vastly more analytical. What can appear in a measurement as a disastrous comb filter, may be heard as innocent spaciousness – room sound, not part of the loudspeaker. This is the fundamental reason why “room EQ” is such a risky business above a few hundred Hz. There is a significant possibility of degrading a good loudspeaker by achieving a “pretty” room curve. At the present time, all of the highest rated loudspeakers, cone/dome and M2, generate room curves that are well described as gently tilted lines at about -0.4 to -0.5 dB/oct from 20 to 20k, in normally reflective rooms and HTs. In spite of the very real differences among rooms, this relationship has been confirmed by numerous independent measurements. If you measure such a curve, it means that you probably bought the right loudspeaker. If you don’t, equalizing what you have to match that curve guarantees nothing. The definitive descriptor above about 400- 500 Hz is anechoic data showing a flat direct sound and relatively constant or smoothly changing directivity. At bass frequencies a steady-state room curve is definitive, and EQ is one of the tools that can help improve it.
Matching sound levels. Years ago I did experiments in which sound levels were deliberately misadjusted. It made an insignificant difference back then, but back then all loudspeakers were much worse than the two you auditioned. Listeners could hear the resonances even with the mismatched levels. In the loudspeakers just evaluated, there are no resonances of any consequence, and minor difference in spectral balance, so what is heard is what is left to hear – directivity, most likely.
Tone controls. High-end audiophiles, in their fantasies, think that tone controls “destroy” sound. You don’t find them on “high end” gear. That could be so if recordings were all perfect. But they are demonstrably not perfect, or consistent in spectral balance, or dynamic range, and so on. Fortunately we humans are very adaptable, and since we have no idea what the final playback in the control room sounded like, we are very accepting of what comes out of our – hopefully neutral – loudspeakers. Still, when I sit and listen to the wide range of music available on Tidal, for example, I hear music with no low bass through to music that was clearly mixed for the bass-deprived masses. Treble ranges from muted to an excruciating screech and tinny transients. I use tone controls. With the best of today’s loudspeakers, recordings may well be the current “weak link”.
So, again congratulations to John and his team, to those of you who believe in science and have taken the trouble to read about it, and of course to the Harman engineers who knew the rules, followed them and created two very fine sounding loudspeakers. It has been an interesting exercise, and we all learned something. :-) Science is forever, opinions change.