A/V Veteran

Late last month I visited the Harman research facility to compare three Revel floor-standing speakers: the Performa3 F208 ($5000/pair), the Performa3 F206 ($3500/pair), and the more upscale Ultima2 Studio2 ($15,000/pair). The venue was Harman's Multichannel Listening Lab, or MLL.

If you're a long-time reader of this and other audio magazines and websites you may have heard about this facility before. To our knowledge there's nothing else like it in the world. Built in the 1990s, it begins with acoustically optimized room, but there's much more to it than that. Positioned behind acoustically transparent curtains is the unique motorized structure visible in the above photo (shown here with the cloth screen pulled aside). When a signal to change speakers is received, the speaker currently in position slides back and the selected speaker comes forward and slides over into the position previously occupied by the first. The entire process, accompanied by the swooshing sound of the sliding palettes, takes roughly five seconds. The acceleration is so rapid that if the speakers weren't solidly secured to their palettes they'd topple over!

You'll notice that only one of each speaker is shown here. That's because at some point after the lab was built (and in contradiction to its name), Harman researchers, led at the time by Dr. Floyd Toole, determined that mono auditioning better revealed a speaker's basic quality. This was important to the R&D process— the major application for such an elaborate, expensive installation.

One problem with mono auditioning is that there's little genuinely mono program material. You can't simply blend the left and right channels of a stereo recording together. Most 2-chanel recordings are produced with spaced microphones, recorded on multiple tracks, and later mixed down to "stereo." This produces intra-channel phase differences that will not combine cleanly to mono. For that reason only one channel of stereo recordings are typically used in the MLL tests, though the facility's capacity for stereo testing remains.

The music for the tests is ripped onto a server in short clips. These are repeated as long as the listener (or listeners) desire, while he or she switches between speakers as desired (the latter can also be done at the tester's volition, depending on the ground rules for the test). When the listener (or the tester) is ready, the test moves on to the next music selection. I retained full control over when to switch to a different speaker and when to move on to the next music selection.

The volume levels of the speakers are matched as closely as possible, which isn't easy to do, particularly when the speakers being compared have fundamental response differences (not the case here). The listener(s) scores each speaker for each music selection, on a scale of 1 to 10, and the results are tallied up at the end of the test.

I provided 15 music selections of my own choosing for use in my auditions. My visit was hosted by Kevin Voecks, Harman's Product Development Manager for Revel, JBL Synthesis, Mark Levinson, and Lexicon, and the tests themselves were run in the blind (to me) by Elizabeth McMullin. On each program selection, the speakers were identified only as A, B, and C. These designations were scrambled whenever the source material changed; that is, A, B, and C were not always the same speaker.

The tests were divided into four sessions, two in the morning and two in the afternoon. The Performa3 F208 came in first in three of the four rounds; the F206 was last in all four. But there was a significant reason for the latter. The less extended bass of the smaller F206 was evident on many of the selections, which kept its scores down slightly. When I later suggested to Kevin Voecks that the F206 might have come in first if enhanced by a good subwoofer or two (from Revel, of course!), while the larger speakers remained sub-less, he agreed. He also noted that there is a small shelf on bottom end of the Studio2 that reduces the level of its bass a bit when compared to the F208 (though the actual bass extension of the two speakers is comparable). Might this have been a deliberate choice to give the Studio2 the "tighter" sounding bass that audiophiles are known to prefer? Or was it merely a re-thinking in the seven plus years since the Studio2's were released to the market?

Despite the bass differences, the three speakers ended up almost tied—not surprising given that they come from a similar design philosophy. The overall average score for all four runs was 7.86 for the F206, 8.3 for the Ultima2 Studio2, and 8.43 for the F208 (rounded off here to two decimal points).

At the end of the day, after the blind sessions were finished, I requested a repeat of five of my music selections. Each were done in the blind, but after I chose my favorite for each of them, I asked to have them revealed, so that I could get a feel for my preferences on different types of music. The results were no further apart. On both a solo guitar piece and an orchestral cut with a strong percussive bass line, I chose the F208. On a percussion track, and also on a closely mic'ed female vocal, I preferred the Studio2. On a more laid back female vocal, with significant artificial reverb, the F206 was my favorite.

Two conditions of the tests may have been significant variables. The use of only one channel of a stereo recording would likely be okay if the listener is unfamiliar with the source material. But if you're intimately familiar with it, as I was, you sense that something is missing, which could be a distraction and skew your scoring. This, unfortunately, may be unavoidable in any such test given the above discussion of the shortage of true mono material. In addition, the listener is not given separate control over the playback level of each selection, other than the overall level overall. A few of the selections were lower in level than I am accustomed to and would have preferred.

But overall this was a unique and enjoyable opportunity to compare three superb loudspeakers, an experience that could not have been duplicated anywhere else.

Addendum: Kevin Voecks of Revel and Harman added some additional details after the above was published. Apparently the volume could be changed with each selection through the same laptop interface I used to change speakers when listening to a particular selection and scoring it. Unfortunately, I was not briefed on this. A software update currently in the works will transfer all user control to a convenient wireless tablet.

In addition, Kevin had the following comments on blind tests using short selections, with which I fully agree:

"As you know, many people argue about double-blind tests. Most of their arguments are without merit, but not all. One of the most important is that in my opinion and observation, it does indeed take extended listening sessions to hear the more subtle differences. The important thing is that these more subtle differences can indeed become more evident over time. Having listened to the Performa3 series and Ultima2 series both for very long periods of time, the difference at high frequencies especially is dramatic. The Ultima2 tweeter is so much "cleaner," with vastly lower distortion (even though the Performa3 distortion is far below most speakers) that it is much easier to listen to without fatigue. Combined with the advantages of low diffraction, it is the high frequency range that causes the Ultima2 series to win in long-term listening tests. Getting back to the blind testing, that kind of difference is best heard with longer sessions. There must be breaks between long sessions, as fatigue sets in, but that is where differences that audiophiles live for become apparent."