[art]by[odo]

It started out of curiosity. And as so many other interests, it took on a life of its own. All I wanted was to find out how many loudspeakers meet a couple of basic standards. I wanted to separate the honest loudspeakers from the pretenders. First, I went through the list of tested speakers and quickly jotted down the ones that passed a couple of basic criteria. Then I got curious about the models that were unfamiliar to me. Then I wanted to know which ones were bookshelf. Then I wondered about the prices. Then I realized that I needed to look more carefully at the measurements to make more objective judgements in order to grade the speakers. The difference between an A and a B had to be clearly defined. Then a closer look made me realize a couple of speakers that I had initially passed didn’t cut it. Then I wanted to see how the bookshelves compared to the floorstanders. Then, well, it went on like this, and on some more. What began as just a little list to satisfy my curiosity, ended up a minor study. Here’s the rest of the story.

The National Research Council of Canada (NRC) does independent testing of loudspeakers, amplifiers, preamps, and headphones. It’s a valuable resource for unbiased data collected with consistent, uniform testing procedures. A survey of the NRC tested loudspeakers was taken for the purpose of sorting out which speakers had better than average fundamentals. The criteria are simple.

1. A flat frequency response, not to exceed +/-2 dB from 100 Hz up for a grade B, 50 Hz up for an A.
2. Low distortion, less than 1% at 90 dB for a B, less than 1% at 95 dB for an A, under 100 Hz excluded.

There was surprise after surprise. Some major manufacturers didn’t make the cut, namely, B&W, Wilson, and VonSchweikert. Yes, B&W. Their top models are known to be excellent. Had those been tested I suspect there would be at least one B&W on the list, but the three models tested were subpar. They certainly don’t put the same effort in their down-line models. Wilson had several models tested, more than any other manufacturer, not one made the grade. There is more make-believe in high-end audio than you can imagine. It’s time to sort the real players from the play-babies.

Other majors barely made it—Dynaudio, Revel, NHT, Thiel—with only one of their models, which doesn’t say much for the rest of their lines. Thiel and Focus Audio just squeaked by. Revel speakers, to be fair, are consistently linear, but most models exceeded the harmonic distortion limits in the upper bass between 100-200 Hz. The Revel Performa3 F206 ($3.5k), just missed out because of that. If price were a factor, I’d have to question what justifies the existence of the Revel Ultima Salon. The F206 looks just as good except for the elevated HD.

The surprises continued with the nearly failing performance of many tweeters—one of the easiest parts to get right. Unknowns made the list, Ethera and Exodus Audio (who?). Mass market producers Paradigm and Energy Connoisseur show up. Paradigm’s Reference Studio 100 is one of only two to get a grade A, and get this, it’s under $4k a pair. Many of Paradigm’s other models are better than average. The first place performer, Magico S5, is the only ultra-hi-end model to get an A. Make a mental note of that. Yet at nearly ten times the price of the Paradigm it ought to shine. The Energy Connoisseur C-9 was a knock-out value at $2.5k, but as with other manufacturers, their other models don’t follow suit, not even close. Goes to show, once again, that neither price nor reputation equates with quality performance. Obviously, hi-end is not so much about performance. It’s more about the fantasy and fairytale of pretend physics and pretty packaging. The Aurum Intergris may be stellar. Its frequency response is, but no HD figures are given. The measurements were left out “because of some technical issues between the measuring system and the amplifier electronics, accurate THD+N and Deviation from Linearity measurements could not be obtained,” according to the NRC. The Aurum system uses internal 300B tube amps. There’s no way to separate the amp HD from the speaker HD. I assume a grade B for those reasons.

Four out of the fourteen are bookshelf speakers. None of them made it down to 50 Hz, not surprising, but neither did the Dynaudio C-4 at $19k, or the Revel Ultima Salon at $22k. Prices range from under a thousand to over forty-thousand. The average, skewed by the four heavy hitters, tipped over $11k. However, the median value brings us down to earth. It’s a “modest” $5.25k. (Don’t make much of this discrepancy between the mean [average] and the median [the 50/50 split point]. It’s just an indication that the speakers exceeding the median have no limit and often exceed it by crazy amounts, in this case by $36,750, the price of a nice car. Also, the accuracy of the prices, gathered from reviews, some over ten years old, may not be current. Still, they allow for a degree of comparison.)

If you think the criteria are too stringent, consider this. It’s drop-dead easy to produce a flat frequency response from 100 Hz up. Less than 1% distortion at only 90 dB isn’t brain surgery. And flat to 50 is not a major hurdle. These two criteria are relatively easy. And they set the foundation necessary for a good loudspeaker. As you can see in the table below, there are speakers meeting those requirements for under $5k a pair, some half again that amount. Surprised?

There was a total of eighty speakers tested. Less than twenty percent met the minimum criteria. Only two get a grade A. That’s less than three in a hundred. Don’t think too poorly of a grade B. The Bs out perform more than four out of five. You should see some of the astoundingly wild frequency responses. A few are worse than in-room measurements. And the worst part of the worst performers is that they are mostly in the several thousand dollars and up price range. Still, frequency response and harmonic distortion don’t tell us everything about a speaker, yet they measure two of the most audible characteristics that have been proven to be the top indicators of sound quality in subjective listening tests with both experienced and inexperienced listeners. Another top criterion for subjective preference is a speaker’s off-axis behavior. This got me looking at those measurements. Half of the fourteen have a better than average off-axis response. But it must be noted, the average in this measure is not good. Better than average only means not as bad as most. This criterion shouldn’t be taken lightly since the same studies stress its importance in the subjective perception of sound quality, second only to on-axis response. Here’s why. If you consider the amount of reflected sound heard in a typical room, usually well more than fifty percent, then you can appreciate the role of off-axis linearity. For details of the studies, read Floyd Toole’s book Sound Reproduction: Loudspeakers and Rooms. In it he cites the extensive testing done by the Harmon Group that found the correlation between objective measurements and subjective preferences.

If the eighty speakers tested by the NRC are taken as a fair sampling of all loudspeakers on the market, it tells a cautionary story. More than eighty percent of speakers are fundamentally poor performers in all three of the most important measures. Less than one in twenty-five get just two of the fundamentals correct. Add in off-axis response and resonances, and those numbers dwindle miserably low.

Plus one more surprise. I recently heard the Revel Performa3 F206. It was set up in the same room as the B&W 803D ($10k). Listened to the F206—they’re good, although there’s not much to crow about. Listened to the 803D with the same piece of music. They sound clearer, more open, more solid in the bass. The superior low end was expected, it’s a much bigger speaker. Then, just to refresh the memory, the recording was played again on the F206. No doubt, and I wasn’t alone in this conclusion, the F206 sounds stuffy in contrast to the B&W. Its treble is good, but the upper bass and lower midrange had a bit of a hand-over-the-mouth quality. Why? I’m not sure. I can, hesitantly, throw out a couple of guesses. My initial thought was some kind of resonance. The F206 is significantly lighter weight and possibly less well internally braced. Other resonances that may be the culprit are the internal resonances exiting through the port. If you put your ear up to a port, you’ll hear a lot more than bass. You’ll hear everything the drivers are producing. And the F206’s port is up front. An easy exit for those resonances, as well as through the driver itself. Yet, the B&W is also ported. The B&W’s port, though, is on the bottom, allowing extraneous port noise to be diminished by diffusion and a less direct path to the ear. There is another factor to consider. The F206’s off-axis is poor—the B&W’s, judging from its crossover points, driver sizes, and baffle, is probably very good. The dullness of the Revels could well be in part a result of the low passed reverberant soundfield created by their very nonlinear off-axis response. Are any of these suspicions valid? These criteria are harder to isolate and measure, but not impossible. Together they could easily add up to significant clouding, blurring, muddling, or whatever term you like for dirtying the sound.

The above was written over a year ago. By stroke of luck, a friend of mine gave me the opportunity to hear the Magico S5 which he had recently purchased. Also, just in the previous week, I heard a couple of other high-end speakers, a $15k pair of open baffle, and a $22k pair of exotic driver 2-ways. Both will remain nameless, both are perfect examples of high-end, high-effort, low-performance wastes of time and money. They’ve put their energies in high production cost and swanky marketing while their lowbrow physics leave their products lacking in every way, poor frequency response, poor bandwidth, poor distortion, poor sound quality. By contrast the Magico are a real treat of well thought-through highbrow engineering. Their name may evoke magic, but they don’t live in the land of make-believe. They get my respect for following the cause and effect of real physics to actually deliver something real for the astronomical dollars they charge.

So, what has Magico done besides the full range flat frequency response, low harmonic distortion, and a very good off-axis power response that makes the S5 sound so smooth and clean? It’s something that explains why the F206 fails. It’s two measurements that are rarely taken, yet something we hear every time we listen to a set of speakers : Internal & External Resonances. Internal are critical because they escape the enclosure through the driver. External are the resonances that conduct through the speaker’s cabinet walls, or in most cases, are radiated by the walls’ flexing (in other words, the walls become secondary speakers). Flat walls are next to impossible to damp or brace. Internal resonances are extreme; remember, the sound pressure on the backside of the driver, inside the enclosure, is as high as outside the enclosure. Reducing these resonances to insignificant values are key. Magico has an almost absolutely dead curved aluminum cabinet. The curve makes for a very strong wall that resists the flex that radiates spurious sound into the room. The heavily damped aluminum scarcely allows sound waves to penetrate. The internal resonances are dealt with, as I understand it from their explanation, by controlling internal reflections (and perhaps through internal baffling to breakup standing waves?). This handling of the backwave explains why the Magico sounds, in a word, clean. It could also be described as transparent, or clear, or smooth, but clean was the word my ears kept telling to me. Mind you, there’s one more consideration. The S5s I heard were being powered by big, muscular amps, almost a 1000 watts per channel. There’s no question that the amps were not being stressed. There were no amplifier distortion products. This is an important qualification, because under powering these large, 4 ohm, power hungry speakers could easily compromise the sound quality.

Upon returning home, while the memory of the S5s were fresh in my ears, I replayed some of the same recordings on my Parallel Audio Project speakers. They have all the attributes of the Magico, full range flat frequency response, low harmonic distortion, excellent off-axis power response (in fact in this measure the Project is superior). They have extremely rigid cylindrical enclosures that do not flex. They sound very clear and clean. They have all the attributes except one. They are not quite as clear and clean as the Magico. Backwave energy transmits through the cylinder wall, although somewhat controlled with a layer of industrial felt, it’s not eliminated. Internal resonances are dealt with by a judicious use of damping material, but it’s only a partial solution. When all else is correct—frequency response, low distortion, power response, and most of the glaring resonances are damped—the final refinement comes down to reducing the internal/external resonances to a bare minimum.

There haven’t been any significant changes to the Project in almost a decade. Now I know where to focus my energy to make the sound of reality more real.

Link to the [SoundStage Measurements Library]
Link to several articles on [Off-axis & Directivity]