Interviews

If you take an ideal line source, there are no floor or ceiling bounces. And because of the cardioid pattern, the radiation at the edge of the speaker is essentially zero, so there's no wall bounce at the intersection with the wall. So with a line-source dipole, you have almost an ideal room match. Reducing the room's effect gives you the purest look at what's on the CD or the record. But this isn't true for frequencies below 100 cycles, where you get the same kinds of room interactions.

That's where servo woofers come in. Several woofers that we and other people have shown over the years have absolute control over the bass. One of the main problems with the bass is that every bass driver, no matter how you design it, is heavy and has a lot of mass. According to Newton's Laws, it takes a lot of force to get this mass moving, and once it's moving, it doesn't like to stop moving. Of course, a woofer not only has to start moving, it has to stop, and go back and forth very quickly. This produces a problem I call "inertial distortion." Inertial distortion is a serious problem for woofers in a box, whether it be a reflex box—and I have very strong feelings about reflex boxes [laughs]—or a direct radiating box, or any other kind of box. There's simply no way the bass will follow the electrical signal. This isn't speculation. This is something you can put a microphone in front of and measure. You can see that there's no correlation with the signal. Not only does it ring when it starts, you can see the length of time before it starts, if you compare the woofer output to the input signal.

There are so many kinds of distortions in a woofer, and the only way to fix them is to put some kind of sensing element on it. We use accelerometers, which allow you to measure every point in that woofer's motion, and the instantaneous acceleration of that woofer. You then can take that accelerometer signal, feed it back to the input of the amplifier, and correct that woofer.

It's like putting feedback around an amplifier. I'm not saying that this is the best thing you can do for an amplifier, but if you do, you'll get 20dB less distortion—conventional distortion. The same thing happens when you put 20dB of feedback around a woofer—you get 20dB less distortion, which is a hell of a lot.

You'll also increase the bandwidth of the woofer such that you can have flat anechoic response. We know that you could get flat anechoic response if you used an accelerometer as your sensing element, and drove the woofer to constant acceleration. Constant acceleration equals constant sound-pressure level, and constant sound-pressure level is anechoically flat. So a servo system makes a woofer have low distortion and flat anechoic frequency response.

There's one more thing, which is probably the most important part of it all. If you design the woofer and servo amplifier correctly, you know exactly how much current it takes to make that woofer accelerate at the same rate as the musical signal. The mass of the woofer is almost zero when you apply the right kind of servo electronics. So, instead of dealing with this heavy thing [the woofer cone] that's obeying Newton's Laws and is way out of sync with everything else—the music and the rest of the system—you have a device that will follow the electrical signal almost perfectly. In fact, with any of our servo woofers, you can put in a 50Hz squarewave and get a 50Hz squarewave out—with the same risetime as the input signal. Now, conventional open-loop woofers [no servo drive] can't even come close to the risetime that is defined by their own bandwidth.

It always seemed logical that the servo woofer was the right thing to do. The first ones were crude, but as we developed them over the years—with cone materials, the kind of motor the system had—they improved tremendously.

One of the problems with servo woofers is that you are making the assumption that what the accelerometer is doing is what the cone is doing. And, of course, with the first cones we used, that wasn't true. The voice-coil would push so hard against the cone it would bend the cone. That caused some problems, because if the cone was doing something funny, the accelerometer didn't know what was happening. If the woofer started wobbling back and forth, the accelerom$*eter wouldn't know it because it only senses perpendicular motion, which you couldn't correct for. So if you try and create a perfect piston in the band, where the cone doesn't break up, doesn't wobble, and can handle the power, then the accelerometer output will reflect exactly what the woofer itself is doing.

That's why we designed the three-layer woofer made of metal, a damping material, and another layer of metal. We also used special spiders and alignment so there is no wobbling. There is one breakup mode at 6kHz, but we cut it off at 80Hz. There are no other aberrations, so the servo system works almost perfectly. That's why, for the first time, Paul and I are happy with the fact that the transition between the woofer and midrange is almost seamless. You can't tell. For the longest time—more than 20 years—you would say, "Well, there's a little difference here." Until now, there was always a reason why the transition wasn't perfect.

It's really happening now. To go back to your question of why I did these unusual things when other people were putting cone speakers together, the answer to that is simple. If you build a two-way, three-way, or four-way speaker—it doesn't matter which—those drivers can be made of compressed horseshit so long as they are a perfect piston in the band in which you use them. You can have a myriad of dissimilar drivers, but so long as they perform as a perfect piston, and you can design a crossover, you should have a seamless, non-sounding loudspeaker. That's a very difficult thing to do, but that's the Holy Grail. People thought the Holy Grail was a full-range planar speaker. But planar speakers aren't perfect pistons—they have their own problems.

With our speakers now, the tweeters are very narrow, have a very specific radiation pattern, and have excellent dispersion. And our midrange is the same way. Our midranges aren't large panels because, at the lowest frequencies, they have perfect dispersion. When they cross over to the tweeter, they start to narrow in, and we're out of them quickly and we don't have the dispersion problems.

McGowan: A big problem was that the electronics often didn't match the woofers. For years, Infinity let [their customers] use [their] own amplifiers. Who knew what the right power amplifier to use was?

Nudell: That's right. Paul and I reasoned that every system we build will have a servo amplifier built exactly for the system—for the accelerometer, for the box, for everything.

McGowan: Once you get the woofer right, get an accelerometer on it, build the electronics right, build the cables right, and build everything as a tuned system, then you can have a seamless transition to the midrange and bass that sounds real. There's a lot to it. Although people have worked on servo woofers for many years, we're just now getting to the point where it's starting to get correct.

Harley: One hallmark of your designs has been wide dynamic contrast. How important are dynamics in reproduced music?

Nudell: Critical. It was driven by my taste in music, which was always oriented toward the large orchestral works—Mahler, Bruckner, Beethoven. Most speakers failed miserably, not only in terms of triple forte, but of pianissimos. No loudspeaker came close. But much of the emotional impact and power of large orchestral music is in those dynamic differences—those are the contrasts that convey much of the emotion of the music. And I always felt that if a speaker couldn't do it—if the system couldn't do it—then much of the emotion of the music would be lost. That's why dynamics are so important to me, and I think why they're important to Paul.

We decided when we took Genesis in the new direction, starting with the highest-end loudspeaker—the Genesis I—that we wouldn't compromise on dynamic contrast. We wanted to make sure that the speakers—if the ancillary equipment could do it—would reproduce the triple fortes.

Harley: You didn't want the loudspeaker to be the limiting factor in the system's dynamic range.

Nudell: Exactly. On the other hand, the other end of the dynamic spectrum was just as hard. Any loudspeaker can reproduce piano, but very few in my and Paul's judgment can reproduce pianissimo and have the same kind of imaging and the same kind of detail. If there's a triangle off to the left and it's triple piano, it shouldn't be some diffuse sound, but a clearly defined image. If a musician hits a music stand, you know exactly where it is. The ultimate resolution of a loudspeaker at the lowest volume has to be just as impressive as the resolution at the highest volume. That was the goal.

Harley: How do you find digital to perform in this area?

Nudell: Just to set the record straight, Paul and I don't find digital offensive these days. It's improving at such a rate that certainly we can listen to music with it and design loudspeakers with it. We like its consistency. We've found that digital will go down to pp—not quite to ppp. The fffs on digital now are better than the fffs on record. I'll argue that with anybody, because I have the master tapes of the record. Digital has come a long way, and I have a lot of hope for it, because it's the wave of the future.

Harley: What other qualities are important to you in reproduced music?

Nudell: It's nice to have dynamic contrast, as you brought up, but there are several other items that are extremely important. Without them, dynamic contrast would mean nothing. In fact, dynamic contrast would be detrimental.

The first is harmonic structure. Every instrument and voice has a unique harmonic structure, and if the speaker or the amplifier doesn't preserve that harmonic structure almost exactly, many of the other parameters are meaningless.

The distinction between an oboe and an English horn in a concert hall isn't very hard to discern. They're both double-reed instruments, but the English horn is slightly bigger and pitched a fifth below the oboe. On most hi-fi systems that don't preserve harmonic integrity, you find that it's harder to distinguish between an oboe and an English horn. They're not distinct enough because the harmonic integrity of the loudspeaker isn't allowing you to hear the harmonic structure of each of those instruments correctly. Preserving harmonic structure is the number-one goal.

Also, this notion of resolving space is extremely important to us. We all know that on some recordings there's a tremendous amount of space between the instruments and tremendous ambience of the hall. Well, that's great, but how do you get it out through the loudspeakers? That's not a simple matter...The loudspeaker has to be able to slice time in very small increments. It has to resolve very small periods of time without blurring one "delta-T" to the next "delta-T." If the loudspeaker can do that, it will resolve the small spatial cues such as sound reflecting from the walls and floor to the microphones.

A clarinet sounds much bigger in a concert hall than the instrument is, because a clarinet reacts with the concert hall. If you don't slice time with fine enough resolution, you're going to blur all those cues that it takes to re-create the ambience. That's why for our speakers we chose drivers that are extremely fast in every frequency range. For the midrange and tweeter we chose ribbons, and we make them out of very, very thin Kapton. That challenge is not only how to make Kapton that thin, but how to put a conductor on it.

Harley: So fast drivers are essential to soundstaging, not just transient response?

Nudell: The whole thing is essential to soundstaging. When I say slice time into small increments, I'm talking about the bass to the top end. If the midbass was smeared from a confused woofer—as most woofers are—that would tend to destroy the natural ambience of a recording and the ability to hear the room the instruments are in. The midbass is critical in that regard. Every driver has to be extremely fast. We've found that the tweeter must respond linearly to beyond 32kHz in order to get the proper risetime that will slice time thin enough in the audible range.

McGowan: Turn off the woofer to your system some time and hear what happens to your soundstage. It's gone.

Nudell: There's a lot of information at low frequencies about the ambience of the room. We allow our speakers to go down to 16Hz, and that may seem ridiculous to some. You get all kinds of noises, but it's those noises that give the feeling of space, the impression that you're really in the room.

Harley: How is loudspeaker design different now from how it was 25 years ago?

Nudell: In a certain sense, there's a sameness to it—that's the art part of it. Throughout the last 25 years, people have asked me to ascribe ratios to what part is art and what part is science. When I first answered that question, I said it was 50% art and 50% science. As time went on, and we got better and more accessible measuring equipment such as MLSSA, I started thinking maybe it was 70% science and 30% art. Now I'm back to the old 50:50 ratio.

Not that we have less equipment now—we have more. We can measure speakers to within a gnat's ass, but the problem we've found is deciding what the hell all these measurements mean. We know what some of them mean, but some of the subtle things we do to high-end loudspeakers really don't show up on any measurements. The scale now isn't fine enough to see some of these very subtle things.

I also wonder what the hell on-axis frequency response means. You don't listen on-axis. What you care about is the average hemisphere of sound that you're going to get in the listening position. Are those speakers putting out spiky ragged crap, or are they smooth from the sides? And if you average them together, will the frequency response be smooth without garbage? That seems to make sense, and relates to energy as a function of frequency. The British seem to think it should fall off at 6dB per octave, but we don't think that. We think energy should be constant as a function of frequency if you can disperse it. Dipoles get around some of that because they disperse the high frequencies enough around the room, and average it better.

Getting back to your question, roughing-in a loudspeaker is a lot easier now than it was 25 years ago. The measurement equipment we had then was very crude compared to what we have now. You try to optimize the measurements, but with every change, you've got to listen to it. There are so many things about loudspeaker design we can't measure. You really need an orchestra in your head to know what's right.

We can rough things in a lot quicker and get a project moving more quickly. Where a high-end speaker once took you two years, now it may only take eight or nine months, because you get that big jump initially.

But understanding all the parameters we've talked about—dynamic range, harmonic integrity, how ambience is re-created, the consequences of drivers not being in time, slow woofers—has to come from experience.

We have much better materials to work with because aerospace always needs new and fancy materials. We can form all kinds of drivers we could never have formed years ago. Sure, we made electrostatics 25 years ago, but could we have made a titanium/silicon-carbide/aluminum dome? No. Could never have done that. Were there materials to make a planar tweeter that wasn't electrostatic? No, because the magnetic materials weren't there, the diaphragms weren't thin enough, you couldn't put the voice-coil on the diaphragm.

I'd say the biggest difference is that we have vastly more material science than we had 25 years ago.

Harley: But that hasn't taken away the artistic aspects.

Nudell: The more sophisticated equipment we get, the faster and easier it is to get to a point, but it won't take you beyond the point.

Personally, I wish it would be more like 80% science—things would be much nicer like that. I'm really more the scientist type. With any piece of electronics, it would be nice to characterize these things that we all hear. It would be nice to know why some cables sound better than others—and there's no question that they do. It would be nice to quantify such things because it would make our lives a lot easier.

It would also bring together two dissimilar segments of our industry. It's a crazy fight we have. We have the logical positivists who say measurements are the only things that matter, and if you can hear stuff you can't measure, then it's a bunch of bullshit (footnote 3). The other segment—which is most of us in the High End—knows things sound differently, and we all hear the same differences, but we don't understand why there are those differences. It would be nice to bring the two camps together. There's no doubt we know more now than we did then about certain things. But do we know more about amplifiers? I don't think we do.

Harley: Do you have a mental image you carry with you that you use as a reference when designing? If so, how important is that?

Nudell: To design loudspeakers of reference quality, it's essential. I can't emphasize it enough. I have the ability to hear an orchestra—or any instrument—at any time in my head. I know exactly how it sounds. The difficult part is to translate what I hear in my head and what I hear from the loudspeakers into a technical reason. Why isn't it closer? That's the big challenge.

But to answer your question, it's essential for any designer of audio equipment to know clearly what things sound like. People ask how you can do that, because Avery Fisher Hall sounds different from Carnegie Hall, which sounds different from the Dorothy Chandler Pavilion. There's no question that all these halls sound different, but instruments have a certain character to them that makes them true. If you know what that certain character is, no matter what hall, it's undeniably that instrument.

A clarinet always has the same character. It may have more projection in one hall than another, and it may be brighter in one hall, but you know that clarinet has that character. If that character doesn't come out of the loudspeaker, then it's wrong.

Harley: I've watched you set up several systems, and it seems as if you know exactly what you want. It's not a question of which amplifier or placement or setup sounds "better," but which is closer to some ideal.

Nudell: First it has to sound right to you. Then you have a chance that other people will like it.

---

Footnote 3: An extreme example of this mindset appeared on the Internet a few months back, where a self-styled "objectivist" stated that, as there was no such measurement that described a stereo image, it was nonsense for audiophiles to discuss the imaging characteristics of audio systems. If it couldn't be measured, it couldn't exist.—John Atkinson