an article with anther interview with Kevin Voeks, this time from the Secrets site. A couple of notable quotes:
|Kevin: Absolutely. Before I was at Revel I was using the Canadian National Research Council because it had the best measurement facilities and Dr. Floyd Tooleâ€™s work there was absolutely best in the field. He went to Harman and continued that great research along with Sean Olive who is now in charge of subjective testing or double blind listening tests at Harman. Thatâ€™s just a priceless resource because the bottom line is how something sounds. Normally listening tests are done in a way where itâ€™s anecdotal and when you do double blind listening tests, you eliminate the variables, and itâ€™s certainly about how the speaker sounds.
Because of having our pneumatic speaker mover, each of the speakers that we are listening to is precisely in the same physical location, so that endless problem that speaker designers have always had about how you compare two loudspeakers technically is completely eliminated. They are in exactly the same place, and they are switched so quickly that our physical aural memory is still intact, which is to the shortest of all our senses in that our aural memory only lasts a few seconds. It means that we are able to do more refined comparisons. You can discern smaller differences reliably when you do the really fast switch. Some people are concerned that rapid switching isnâ€™t their style--they like to listen for extended periods. The way we do listening, you can do it anyway you want. If you want to listen for half an hour before you switch, thatâ€™s fine. Thereâ€™s no pressure to do it otherwise in the Revel listening test.
The ability to do accurate measurements is really critical. Thereâ€™s no replacement for a large real anechoic chamber. Basically the industry is using pseudo-anechoic measurements that are achieved electronically, and they donâ€™t work in that they throw away so much resolution depending on how large the area is in which you are measuring that you miss critical data. Not only do you get inaccurate low frequency measurements, but you also have problems with not being able to see resonances, or evidence of resonances.
|Sumit: How well do the measurements correlate with the actual blind tests?
Kevin: Thatâ€™s a perfect question. Because we actually have come up with measurements based on listening tests, itâ€™s a full circle. We donâ€™t do measurements because we have cool equipment to do it, we do them only because we found an audible characteristic that will differentiate two speakers or one thatâ€™s more refined.
So how do we measure that? How do we quantify that so that we can really engineer the speakers to perform better in that regard. So thereâ€™s extremely good correlation between the measurements that we do, the set of measurements, and the sound quality. Looking at a set of measurements, if theyâ€™re not good, we can say absolutely for sure that the speaker will not excel in listening tests. The converse is not completely true. Looking at a good set of measurements we could say this is a good speaker, but the difference between a really good and really great speaker is still beyond our ability to absolutely predictably measure. So itâ€™s that last level that we spend a lot of time on because itâ€™s not as direct a path.
Note that I provided the italicized & bold text to make a point.
And with regard to the priorities in accurate sound:
|Sumit: I find that quite commendable. When designing a loudspeaker what are the three most important aspects of a loudspeaker according to you that one must get right?
Kevin: Timbre is the overwhelming aspect. Based on our blind listening tests timbre is the thing that differentiates between good and bad loudspeakers, but also between good and great loudspeakers. So timbre is kind of a broad term. It incorporates balance, frequency balance, or it can be thought of very roughly as frequency response. Thatâ€™s a little dangerous because, not to infer that one on-axis measurement tells you what the frequency response is in a loudspeaker. It doesnâ€™t. Other areas like off-axis response are very critical and weâ€™ve learned that very far off-axis response like 60-75 degrees is very critical. Almost no one even measures it, let alone designs loudspeakers that are optimized at that sort of angle. But weâ€™ve looked at real world situations and found that the all-important side-wall first reflection is a function of the speakerâ€™s output at that kind of angle in the vast majority of listening rooms. So it means that youâ€™re going to be hearing that kind of sound. You will hear it with a slight delay, and in many rooms without very much attenuation. So optimizing the response at that kind of extreme angle is very, very important. And then the power response, the reverberant field that we hear a little later in time is also important. So we literally design for all of those areas: the direct sound, the first reflection sound, and the reverberant field, because we know that all those three things are huge contributors to the timbre, to our perception of the speakerâ€™s timbre.
Finally, with regards to time/phase (I know this wasn't brought up, but it's a valid queston as to why one measurement can be more indicative of sound quality than anohter:
|Sumit: You mentioned high-order crossovers a while back. Revel speakers have always used high-order crossovers. Several companies take the opposite approach i.e. they use low-order (first and second order) crossovers. What are the disadvantages of using low-order crossover designs? Are there any advantages to using low-order crossovers?
Kevin: Well, we were fortunate enough to have done research that has allowed us to know, to really understand, what are the characteristics that are important to sound quality and what are the characteristics that have some value but less, and those characteristics that donâ€™t have any sonic value and that allow us to make the optimum choices in the design. That all points very, very strongly to high-order crossovers because high-order crossovers are necessary in order to have low distortion which is way up there on the list of important sonic qualities.
High-order crossovers are important to have good dynamic capability without compression. It would really shock audiophiles to see how much the response of most high-end loudspeakers changes at different volume levels. They are like completely different loudspeakers when played even at moderate levels, and it is something that is very directly measurable. So we really focus on making sure that not only is the timbre really accurate, but that it changes as little as possible over a very wide dynamic range. Plus the distortion is below the audible threshold; resonances are below the audible threshold because our research has shown those are really important things.
If we used first-order crossovers, we would degrade the off-axis response, and therefore the timbre, we would completely degrade the distortion characteristics, we would loose our dynamic capability, our freedom from compression because tweeters and mid-ranges are then getting signals that are outside the frequency range that they are really designed to handle. So itâ€™s really mostly heat, and that heat makes the voice coil impedance go up, and as a result of that the filter network is mis-terminated because itâ€™s not seeing the termination impedance it expects to see, and then the response of the crossover is impacted.
I think the bold text I quoted explains my position perfectly. I don't just expect very expensive speakers to have relatively flat response, I expect anything that I am going to buy to have that, short of maximum extension (obviously within reason).
Here are three graphs (all done by Stereophile) of Revel speakers on three different occasions. The price points were (at time of test) 3.5K, 10K, 18K (or so). This in my opinion is the proof of that what Kevin Voecks was talking about in those interviews is actually caried out in the product he is responsible for:http://www.stereophile.com/images/ar...t/revf3004.jpghttp://www.stereophile.com/images/ar...t/Rusfig04.jpghttp://www.stereophile.com/images/ar...rt/Salfig4.jpg
The graphs are in order from least to most expensive. I am looking forrward to see if they can keep up this kind of performance in the upcoming Stereophile review of the Revel F12 (Kal's got a set coming).
Obviously there are lots of ohter examples. The graphs I've seen of NHT, Energy Veritas, Aerial (for the most part); and Triad, among many others, all try to atain flat FR as one of their design goals. Just to prove that this isn't a Revel add, here's an exerpt from a Keith Yates In-Wall test of a Triad system (sorry, no graph :p ):
|The success of the subjective testing was duplicated in the objective analysis which followed a few days later, disclosing for the benefit of the eyes what the ears had already sorted out: a textbook-right frequency response through the all-important midrange. From about 78Hz to a little over 1kHz, a space of four of our hearing system's ten-octave range, the on-axis response fits within an extraordinary +/- 1.5dB windowâ€”this without the benefit of any statistical smoothing. (Adding one-third octave smoothing, something rather akin to what our hearing system does naturally, lets the curve fit within an almost unheard-of +/- 1.3dB window.)
Likewise, the off-axis average shows exemplary control, confirming that the speaker ought to be considered a leading candidate for installations necessitating placement in less-than-ideal locations relative to the main listening area. Priced at a fraction of B&W's stunning new AWM70, and Triad's own equally stunning InWall Gold system, which has defined the performance possibilities for the whole in-wall genre for several years now, the OmniPlus has quietly emerged as the odds-on favorite for those who care deeply about music but don't want to spend thousands on in-wall speakers that do it justice.
The full review (along with several other in-walls) can be found here.
If measurements have no correlation to how accurate a speaker sounds, why do them with subs? Why not just say: "this sub sounds great with movies and music; and can play real loud and real low"? Wouldn't that be good enoungh? Why do you think Ed Mullen has added additional measurements to his sub tests, to be sure to record every possible variable, etc; if they aren't important? Do you think he's wasting his time in his efforts?
Bringing in examples of expensive high-end stuff which meausures poorly is easy, as a lot of the high end is completely catering to the "subjectivist", who claim to be able to hear every nuance in cables, etc. Most people who buy this stuff could care less how it measures. It's like art to them; many aren't trying to trully re-produce anything. Take a look at the measurements of an SET amp, for example? Talk about low value spec-wise :D.
Specs to me are important. So is listening, but since I can't listen to everything, I look at specs to decide what to audition. And, for a variety of reasons, I find the "only subjective reviews" to be of much less value to me.