Interviews

Norton: I mentioned, in passing, in a recent article that the listening room is the last frontier of audio. You've done some interesting things in this area. The other day you mentioned to me a new program that Snell is developing to help the customer select the right listening position and the right position for his loudspeakers.

Voecks: That's right. It calculates all the room modes based on its being a rectangular room, then predicts the areas where you will activate room modes and where you won't, and gives options from best to worst for speaker placement. That will be a free service for Snell owners.

Norton: Do you have any particular feeling with your loudspeakers, or personal preferences, for either a live listening room or a more heavily damped listening room?

Voecks: As you can see, this room, where I do a lot of listening...

Norton: ...is really live (footnote 6).

Voecks: It is. It has overstuffed furniture, but there's no Sonex on the walls. There are no Tube Traps in here right now. This is the way I usually do the listening because it's like a normal room. I talk to a lot of customers every day who have questions about setting up their systems, and it's obvious that very, very few of them are willing to significantly treat their rooms. They usually can't get away with it, in a domestic situation. So most of my listening is done like this. I also bring in Tube Traps and Sonex, and do listening and comparisons under those varying sorts of acoustical conditions. The IEC room up at the NEC is very, very neutral. They have bass traps to make sure that it's also flat at low frequencies, and the positions that the listening chairs are in and that the speakers are placed are ones that are pretty much equally affected by the room's acoustics. So its a neutral situation.

Norton: Have you had any experience with the Live-End–Dead-End listening room that is promoted in some quarters?

Voecks: Yes, I have set up Live-End–Dead-End rooms in a couple of different rooms and done a lot of listening. I don't think it's the most neutral environment. I can see why it has gained popularity in recording studios, but I think that it degrades the reverberant field. Peter Snell always said that the early arrival sound had to be ultra-ultra-flat—and that's what you heard in terms of imaging and specificity—but that the reverberant field also had to be carefully controlled, and if you had a dead-end situation you're going to roll off the top end of the reverberant field. I don't think that's desirable.

I think it's amazing that most speakers are designed really without thinking about the room. The Type A, in its first version, totally accounted for the room's boundary effects. In that case, the woofer was located within acoustical contact of the floor throughout its range, so that there were no peaks and dips caused by that reflection. The midrange was kept well away from the floor, and there was absorbent material below it which reduced the first reflection off the floor. The curved baffle made it act as if the speaker was in a bump in the wall—an infinite baffle. We are continuing to pay close attention to how the speakers act with the room. For instance, the Type B, which we showed a year ago just to demonstrate this concept, uses two woofers which operate over the same frequency range but are located at different distances from every boundary in the room—the floor, the walls, the ceiling—which minimizes the frequency-response irregularities that you otherwise get with a conventional speaker.

Norton: I recall you mentioning once that in the design of a loudspeaker, you do a great deal of your listening in the mono mode, vs using a stereo pair.

Voecks: That's correct. Research has shown conclusively—again this is NRC research—that you come to the same conclusions about which speaker is better whether it's in a stereo or a mono mode, but you come to that conclusion faster and more reliably when it's a single-speaker comparison. It gets terribly complicated to switch between stereo pairs and keep their interaction with the room relatively the same—that's a big problem in itself. Plus when you're working on a prototype, having to continually change two of them would slow things down. So most of the original listening is done with a single speaker. I don't use an actual mono mode [combining the two channels of a stereo signal] because that can, depending on the recording technique, roll off the top end and give you some false results.

Norton: Do you use dedicated mono program material or one channel of a stereo signal?

Voecks: I use one channel.

Norton: You mentioned the Type B in passing. I don't know how much public information you want to reveal as to where that model stands. I know it was mentioned quite favorably in some CES reports about six or eight months ago. I'd guess that final development and production are still quite a ways down the road?

Voecks: That's right. It's the first model I've ever done that has been on schedule, although it has been a very much revised schedule from the first. We will be showing a completely finished version of it at the January 1990 CES.

Norton: I know that it's probably not even in concept yet, but is the original configuration of the A considered more or less enough of a classic that you'd want to retain that in any future revision, or would you consider going to a completely new type of design?

Voecks: It's a little early to say. Certainly the curved baffle is a great idea. I've done a lot of work on diffraction effects and how audible they are, how measurable they are, how to solve them, and how to get rid of them. The CII and the Q, for instance, have a grille frame that is absolutely flush with the baffle, and it then has a gentle curve. That's very effective. The closer to the driver there is any irregularity in the surface—particularly to the tweeter—the more there are diffraction problems. And diffraction problems are real amplitude-response problems, both on-axis and in the alteration of the reflected, off-axis sound.

Norton: All Snell loudspeakers are, obviously, direct-radiator loudspeakers. It must be somewhat frustrating for you, as a designer trying to push the boundaries on direct-radiator loudspeakers, to realize how popular dipoles are in the audiophile market.

Voecks: That is frustrating. Personally, I believe that tonal accuracy is by far the most important thing in speakers, and certainly the blind tests that have been done have proven that to be true. However, in our very subjective field, people often are willing to completely trade away tonal accuracy for a great sense of depth. It could be argued that in a dipole or bipole loudspeaker this is clearly produced by the reflection off the rear wall. It is not reproducing something in the program material; that is simply impossible. It has to be producing it by that reflection. But the argument is, so what? If, on orchestral material usually, that sounds more realistic, then that's OK. I don't have a problem with that. People usually are very clear about which they like. There's not that much competition, really. Most of the stores we sell through also sell dipole speakers, and they acknowledge that customers usually prefer one or the other. They're usually not torn between them. That was the idea of the Mirage M-1: what happens if you made a good dipole speaker—one that has accurate frequency response?

I asked Kevin at this point if he had any comments he'd like to make that weren't triggered by the questions or discussion. He added some further elaboration on his and the NRC's measurement techniques, and on his crossover optimization procedures.

Voecks: Amazingly enough, most manufacturers measure the frequency response of their speaker at wherever it looks the best; they adjust the microphone to that position. It is much more logical to look at a window—let's say a ±15° window—where the on-axis listeners will be, and average those spaces. We look at that at the NRC facilities. We also look at what's going to contribute to the early reflections. The first major thing that we hear, of the two major sonic characteristics of speakers, is the early arrival sound that's composed of both the sound that is launched directly from the speaker to the listener, plus nearby reflections—usually the floor and a sidewall are the dominant early reflections. Your ear-brain combination cannot separate those two events because they are too close together in time.

You can see, obviously, that the content of that reflected sound is very, very important because it's being added together with the direct frequency response. That's usually not explored at all by speaker companies. We're looking at this off-axis response, both vertically and horizontally, including clear out to 70° off-axis, to make sure that the response is smooth and doesn't have huge peaks and dips in it that would color the sound.

The second big thing that we hear is the reverberant field. At the NRC we're able to measure completely around the speaker in every direction, to see what it will contribute to the reverberant field in a room, in total. You can also optimize that. High-order filters contribute greatly toward getting that optimized. After we do all of these important, off-axis measurements and have this great wealth of data from the NRC, then I use a filter simulation program to work on the filter networks. One of the reasons that so many audiophile companies have concluded that first-order filters are better is that they can't implement properly designed, high-order filters; if they've tried, the filter has been a failure. It's certainly not a trivial task. On the Type CII, for instance, I spent about 1500 computer hours getting it really flat. That's just not something you can do by the seat of the pants. You simply must use computer-assisted design.

Norton: Obviously you have found that the computer aids that are available now either speed the process or make it more efficient.

Voecks: They also allow you to do much, much better work than you could possibly do otherwise. In one minute you can try a completely different crossover network, a different set of drivers, move the listening axis, or change a driver. You used to have to actually build all of them.

Norton: Where do you see the future of loudspeaker development going from here?

Voecks: Recognizing that the marriage of audio and video is finally becoming a reality, I see genuine hi-fi speakers which are also specially suited for use with Dolby film sound. This would apply to both the front speakers and the specially optimized surround speakers. The adoption of a minimum performance standard would go a long way toward an experience in a home listening room which is what the director intended. I also believe that as more and more people experience objective listening tests, they will start to pay more attention to the things that really make important audible differences.

After the interview, I had a chance to listen to both the new Type EIIIs and the early prototype Type Bs (the same versions—actually the exact same systems—that were demonstrated at the 1989 Winter CES). Kevin used some of the same program material he has used in his show demos: CD selections dubbed onto DAT. The Bs sounded as promising as I remember them, but are apparently to undergo a significant transformation, both in drivers and in cabinet, prior to their next showing (and hopefully their formal introduction) at the 1990 WCES.

Kevin also demonstrated his computer design tools. He uses three programs: MLSSA (Maximum Length Sequence System Analyzer) from DRA Laboratories for quasi-anechoic measurements, LEAP (Loudspeaker Enclosure Analysis Program) from CNS Electronics for low-frequency alignment, and the Loudspeaker Crossover Design and Optimization Program from Peter L. Schuck Consulting. All of these programs are available on the open market and all were, as an interesting coincidence, advertised in the 4/89 issue of Speaker Builder magazine. They are all rather expensive for a casual speakerbuilding hobbyist, but only MLSSA (at $2750, including the A/D board) would be prohibitive in such an application. (The other two programs are about $200 each.)

The quality of the results obtained from any of these programs is, of course, critically dependent on the inputs provided by the user. The sophisticated crossover optimization program, for example, requires that you input both the driver frequency responses and impedances. If you use the driver specs from the manufacturer for these inputs, you'll get a result analogous to measuring something with a micrometer that you've just cut with a hacksaw. Kevin Voecks, of course, can use the precise data obtained from his measurements at the NRC, which makes the precision of these programs worthwhile.

---

Footnote 6: KV's listening room, where the interview took place, was of moderate size, with a standard 8' ceiling. It was carpeted, but the windows were undraped (blinds only).