Reference

Revel began demonstrating prototypes of the Performa F228Be, reviewed elsewhere in this issue, at the Consumer Electronics Show and other audio events in 2015, though they had yet to settle on the model name. The most salient feature that distinguished the F228Be-to-be from the established Performa3 F208 was the new beryllium tweeter, and while it sounded more than okay under show conditions, I always heard a bit of brightness and harshness. I knew it was still a work in progress, but I wasn't very eager to want to take it home.

My thoughts began to change after an old friend consulted me about buying some new speakers. After hearing my suggestions, he bought used pairs of his four top candidates and set them up in his apartment. He reasoned that it would be impossible to find a single dealer at which he could compare them side by side, no series of in-store demos would ever tell him as much as hearing the speakers in his own listening room, and he could buy them used and, after choosing one, resell the rest with little or no financial loss. The opportunity to compare, side-by-side in a comfortable domestic space, four well-known, highly rated floorstanders, all costing between $5000 and $10,000/pair? I couldn't resist.

He and I and another friend spent an evening listening to familiar music through all four pairs of speakers in turn. It wasn't a blind test, and going in, we'd each expressed an expectation of a preference for one of them. There was no order to our testing—we listened to a pair as long as we all agreed we wanted to, taking turns sitting in the sweet spot, then unhooked them and moved in the next pair. Back and forth we went, track after track, speaker after speaker, discussing what we heard as went along, until we'd reached a consensus.

I won't reveal the identities of the three other speakers, as the results of our very informal "test" are probably relevant only to the three of us. We share a similar perspective on audio; we listened only to classical music, mostly vocal and orchestral, and heavily late and post-Romantic (Brahms, Mahler, Rachmaninoff, etc.); and our friend's drop-dead view of the Manhattan skyline could not have failed to enhance the subjective impressions of the speakers that framed it.

As I said, our choice was unanimous: the Revel Performa3 F208. I tell this story because the F208s sounded so much better in this room and system than I'd ever heard them, whether in stores or at shows. When the Peforma3 F208 was launched, I'd been disappointed that Erick Lichte was assigned to review them. Now I was downright peeved.

I mulled these results, wondering how I could apply them to my own situation as reviewer and audiophile. Two pairs of speakers at a time is the most I can accommodate in my listening room, and swapping them into and out of their optimal positions is difficult at best. That's why I make most of my comparisons with long auditions of a single pair, supplemented by detailed notes suggesting what to attend to when finally switching to another pair. But there's always the nagging annoyance of not being able to scratch one itch: With the sound of one speaker fresh in my mind, I want to hear right now how another speaker does this!

Well, Harman International's Multichannel Listening Lab contains a way of doing this instantly and effortlessly: the latest version of their speaker mover permits blind, monophonic comparisons of loudspeakers (footnote 1). As I'd just recently installed the Revel Performa F228Be's in my system for the review, I leapt at an offer to visit Revel's Acoustics Technologies Manager, Kevin Voecks at Harman's facility in Northridge, California, and to subject myself to the rigors of comparing three speakers—one of which, I assumed, would be the Performa F228Be.

In addition to strenuous sessions of blind testing, the big day included visiting other Harman and Revel staff. First, I was given a primer in the design of transducers from Principal Engineer An Nguyen, who emphasized how Harman is able to optimize transducer parameters with computer-aided design (CAD)—most potential problems can be anticipated and resolved before the first prototype is built. Afterward, engineer Larry Brown showed me how he incorporates the design parameters for all the components into a comprehensive specification document. That design-spec doc is interactive—as engineering parameters evolve through further testing, individual changes percolate through it, and all related and dependent components are accordingly updated. Finally, I had the pleasure of reconnecting with Principal Engineer, Mark Glazer, who guided me through the acoustical and mechanical analysis of the drive-units, as well as the acoustical analysis of complete systems.

One of the important results of that analysis is the generation of what Floyd E. Toole, Harman's former corporate vice president of acoustical engineering, has dubbed Spinorama graphs, which he presents as highly reliable predictors of accurate and satisfying audio listening (footnote 2) What I came to appreciate was how the Spinorama data, particularly the listening-window response (±30° horizontal, ±10° vertical), predicted what I heard in the listening tests.

Harman documents Spinorama results not only for their own speakers but also for speakers made by many of their competitors; other companies, such as Bryston, now offer on their websites listening-window and total-sound-power frequency-response data, along with the traditional on-axis response. By subtracting the listening-window data from the total-sound-power data (or from the early-reflections data), one can derive "directivity index" (DI) curves that reveal the consistency of the off-axis frequency response. The latter don't play a major role in Harman's Multichannel Listening Lab, which compares single loudspeakers (interactions between the speakers of a stereo pair are not involved) placed far away from room boundaries (early reflections play no significant role). However, a smooth, consistent DI curve should indicate greater ease in positioning the speaker in a room, and better in-room sound.

These presentations helped me better understand what goes into the creation and manufacture of Revel and other Harman loudspeakers, and served as respites for my ears and brain between fatiguing audio trials. They also gave senior acoustic engineer Omid Khonsaripour of HarmanX, Harman's research group, enough time to randomize the order of the test speakers for each trial.

The setup: From a list of familiar tracks/albums I'd sent earlier, Voecks selected 10 excerpts of less than 30 seconds each, and loaded a mono channel (left or right) clip from each into the Multichannel Listening Lab's control system. I was told that behind the curtain in front of the listening seat were three loudspeakers, designated A, B, and C. Using a tablet, I could 1) select a track to play, 2) go forward or back within the track, 3) adjust the volume, and 4) select among Speakers A, B, and C.

Because all of these options were available all the time, I could switch speakers within a track, though this interrupted playback for about three seconds as each speaker was actually and automatically physically moved to the same position behind the curtain. At first, I used three of the tracks to get a handle on how I felt about the three speakers, switching among them in each of the six possible permutations. Then I did back-to-back comparisons of the three possible pairs. By then, it had become clear to me that the speakers did sound different, but that some differences were easier to hear than others.

Thus began the obsessive phase. To define the differences, I selected a particular phrase of music, instrumental or vocal, and repeated all of the above comparisons for each. When my notes indicated a consistency of description of the characteristics of the candidates, I realized that I'd been sitting in semidark, focusing my mind on snippets of sound, completely disconnected from reality and the passage of time. How long had I been doing this? Though I felt I'd been doing this all day and into the wee hours of the next morning, it turned out that only 45 minutes had elapsed. I concluded the session by listening to each track straight through with each speaker to confirm the impressions I'd noted down, then bolted from the room to reconnect with the real world.

After lunch came another session with another subset of musical selections, and in late afternoon a third session. During each of these I followed the strategy I'd used in the first. I wasn't informed about which speakers Voecks had chosen, or whether or not Speaker A was the same speaker in all three sessions. I was pretty sure that C was always the same, but I also thought that A and B were swapped for Session 2. As it turned out, the letter assigned each speaker remained unchanged throughout all three sessions.

Who were those masked men? Speaker C was not a Revel, but a speaker very similar to one I'd reviewed, quite favorably, in Stereophile. Its bass reproduction wasn't problematic, but in direct comparison to Speakers A and B it stuck out for its notably recessed upper midrange. Its Spinorama graph (fig.1) also reveals an elevated HF; though I hadn't noticed that as such, it may have contributed to the deemphasis of MF presence.

Speaker A, it turned out, was the Revel Performa F228Be (fig.2), and Speaker B was Revel's top model, the Ultima Salon2. Both seemed open and balanced, with a detailed but subtle treble, but while some tracks favored one or the other, it was ultimately a standoff. I'd thought that Speaker B had a tighter, less prominent midbass, and had guessed it was the F228Be. But Speaker B was the Ultima Salon2, and in nonblind post-test listening it became apparent that its cleaner midbass was coupled with a more satisfying deep-bass experience, especially at high volumes. Nonetheless, it was on that basis that I'd chosen the Salon2 as my preference in Sessions 1 and 3—but had chosen the F228Be in Session 2.

These results surprised me. Despite their huge differences in price, size, and vintage, the two Revels were so similar in sound quality that I couldn't reliably distinguish them under these blind conditions. I'd fooled myself into believing that Speakers A and B had been swapped for Session 2, then swapped back for Session 3. Perhaps more sessions would have revealed a statistically reliable preference—but no way will I subject myself to that.

I came home with ears and mind refreshed, and certain that any thinking audiophile would benefit from the experience. That said, from now on, good-looking Spinorama graphs or similar data (footnote 3) will be major criteria in my considering what I'll review—and what I'll buy.—Kalman Rubinson

---

Footnote 1: See Thomas J. Norton's description and experience of listening to Revel speakers in the Listening Lab here.

Footnote 2: For a full description of the development, application, and use of Spinorama graphs (Chapter 5), and Floyd E. Toole's insights and overview of acoustics and loudspeakers, I highly recommend the third and latest edition of his book, Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms (New York: Focal Press, 2017).

Footnote 3: I have thought about duplicating Harman's Spinorama graphs for the measurements that accompany Stereophile's loudspeaker reviews. However, for consistency with the reviews published since 1989, I have stuck with the responses averaged across a 30° window centered either on the tweeter axis or on the listening axis recommended by the manufacturer.—John Atkinson