Reprinted with permission from Stereophile Magazine, Vol. 19 No. 4, 1996.

Interview

Stereophile interviews Siegfried Linkwitz

By Shannon Dickson

Siegfried Linkwitz was born in Germany in 1935. He received his electrical engineering degree from Darmstadt Technical University prior to moving to California in 1961 to work for Hewlett-Packard. During his early years in the US, he did postgraduate work at Stanford University. For over 30 years Mr. Linkwitz has developed electronic test equipment ranging from signal generators, to network and spectrum analyzers, to microwave sweepers and instrumentation for evaluating electromagnetic compatibility. He has also had a long and distinguished second career as an audio engineering visionary. Along with Russ Riley he developed the famed, and widely used, Linkwitz-Riley crossover filter in the mid-1970s. Since then he has contributed several important technical papers covering a variety of measurement and speaker issues to such publications as the Journal of the Audio Engineering Society, Electronics (Wireless) World, and Speaker Build.

Most recently, he has joined forces with fellow HP engineer Marshall Kay, CAD (computer-aided design) specialist Kurt Pasquale, and marketing consultant Tom Hoffman to form Audio Artistry. This three-year-old, North Carolina-based company is dedicated to developing and crafting speakers based on the accumulated insights and wisdom Mr. Linkwitz has gained over three decades of loudspeaker research. I spoke with Siegfried about some of these insights and experiences during the course of evaluating the Audio Artistry Dvorak, the review of which is found elsewhere in this issue. My first question concerned what had motivated Linkwitz to get involved in audio.

Siegfried Linkwitz: I grew up in a family in which music was very much appreciated. My father and brother played the piano, and although circumstances during World War II prevented me from learning an instrument, I've always had a love for music. After graduating from university and joining Hewlett-Packard to design electronics, it was only natural that I wanted to build audio equipment I could use at home, so I got very involved in building power amplifiers, FM tuners, preamps, and you name it -- anything electronic I needed to reproduce music. Then I had the fortune of meeting some other engineers at HP who were similarly involved in audio, particularly Lyman Miller and Russ Riley. Lyman was very much into electronic design and making recordings while Russ built amplifiers and had a keen interest in speaker development. They really turned me on to investigating things even deeper, and loudspeakers, to us, were the most interesting and challenging area since so little was really understood about them. The speakers then on the market could certainly be improved, so we saw a real chance to make a genuine contribution.

Shannon Dickson: Could you share with us some of the fundamental problems you and your colleagues encountered during the early attempts to improve speaker performance?

Linkwitz: One of the problems at the time was that good test equipment wasn't available to us. Russ Riley developed his real-time 1/3-octave analyzer and a pink noise source which we used to make in-room measurements. I bought an early Advent speaker, measured it using the real-time analyzer, and consequently developed an equalizer to flatten-out its frequency response. That was a first attempt on my part. I then experienced a real surprise after we went to some local stores and heard the Electrostatic Sound System's ESS-7. It just sounded great, much better than the Advent. Naturally, I bought the speaker and took it home, but after measuring it, I was astonished -- it measured very poorly! That led to a whole investigation into why it sounded so good but tested so badly.

We found out rather quickly how important driver quality was, as well as the distortion contributions of cabinet resonances. We began experimenting with wool stuffing in the box and various bracing and panel damping techniques. We found that wool could be a very effective loading material. A number of commercial designs sounded much better when we replaced whatever they had inside with natural wool fiber. In my early designs, we tried two basic concepts built around rather small enclosures, both of which worked quite well. For instance, we made some very rigid, heavily braced small monitors; then we went the other way, using very limp, thin panels for the box construction. These were very easy to damp by applying roofing tar with sand mixed in. As you can imagine, this was a real messy operation -- it smelled pretty bad too, particularly if you placed the speaker in the sun. It would out-gas for several weeks before you could tolerate the smell!

While it damped box resonances quite effectively, this approach was not really practical from a commercial point of view, nor would it have been a very welcome addition to most people's living rooms. But it did demonstrate how important minimizing box resonances is and just how difficult it is to really control this form of resonant behavior.

Dickson: You've worked with some of the most respected engineers in audio over the years. Who had the greatest impact on your thinking regarding speaker development?

Linkwitz: I mentioned Lyman and Russ already. Lyman was really into the recording side of things, so he did a lot of recordings on a semiprofessional basis and was particularly interested in capturing sounds as close to their natural origin as possible. So we had some great reference material to guide our evaluation. I learned a lot about recording from Lyman and continued to make many of my own reference recordings, which I used extensively during the development of these new speakers. Russ Riley is a very ingenious design engineer and, on top of it, a superb listener. I was always impressed by how easily he could identify just what the problems were in a speaker and in what frequency range and what one needed to do about them. He had absolutely superb hearing. While not as well-known as some of the other engineers, both Lyman and Russ had a big impact on my early audio career.

Through my work in developing test equipment for Hewlett-Packard, I met Laurie Fincham [then with KEF, now with Infinity] and we became good friends. We've shared a vast amount of information with each other over the years, have met frequently, and consequently had some very positive mutual influence on one another. Through Laurie, I was also introduced to a number of distinguished engineers such as Floyd Toole, Stan Lipshitz, John Vanderkooy, and Peter Walker from Quad. I had been following all of these people's writings very intensely all along, so it was a joy to meet them.

Most of these folks have been at my house at one time or another to listen to various ideas I had been working on. In addition, I have been an avid reader of the JAES throughout the years, as well as Wireless World from the UK [now Electronics World -- Ed.]. Wireless World used to carry a great deal of high-quality information about audio and speakers; it still does, in fact, though it's not as easy to find these days. Actually, my first publication appeared in 1978 as a lengthy three-part article in Wireless World in which I described the construction of a three-way active speaker system consisting of small satellites and a subwoofer.

In summary, the various influences on my thinking have led to a general approach that is really a blend of the analytical -- meaning the measurement of things -- and the subjective listening experience, to try to find out what is really going on. If there is a hypothesis of why something works -- this way or that -- I'll set up an experiment to see if I can prove it or disprove it. In this way, I've always attempted to correlate what we hear with objective measurements -- not always successfully, mind you, but at least making the connection where possible. This method will give you a lot of insight into which measurement or artifacts are important and which are not so important. Occasionally, I've found results that look very significant on paper but are barely perceptible, if at all, while on the other hand, some extremely slight irregularities can by very important sonically.

Dickson: Can you tell us what your priorities are in making and evaluating specific measurements?

Linkwitz: I've learned there is a whole battery of measurements one needs to use -- and interpret correctly -- in order to get a better picture of any given speaker. No one measurement will tell you the whole story. At the top of the list is definitely a loudspeaker's on-axis anechoic frequency-response measurement because this represents the direct sound you hear. However, of similar importance are the vertical and horizontal anechoic off-axis responses. So in my designs, I try to achieve a very well-behaved off-axis response which duplicates the shape of that on-axis, but steadily decreases in level the farther you move off-axis. This is so important in determining the reverberant field and the reflected sound in the listening room.

Another key factor I learned during the development of my crossover design is that, when two drivers are combined in the crossover region, their summed output should be at its maximum on-axis. In other words, the radiation pattern remains stable at the crossover region and doesn't shift. For example, I've found through experimentation that it is definitely audible if you go some distance above-axis and all of a sudden have a maximum peak or sharp dip in the crossover region. This problem is similar to what happens with many large-panel dipole designs. As they produce higher frequencies, their off-axis response becomes more irregular, with peaks and valleys that can color the overall sound and make speaker placement in a given room very critical. If the crossover on any speaker doesn't blend together, you can get this kind of off-axis peak.

Another measurement I look at is the overall frequency response on a half-octave or octave basis, just to see the general trend: whether the treble is rising or sloping, etc. When you look at any response in detail, you never get a flat picture, you always have little ups and downs; but I've found you don't really gain anything by trying to smooth out these small ripple effects in the response. However, how smooth the response is over a third- or half-octave basis is important. I'm essentially looking for an averaged-flat anechoic response.

I do my quasi-anechoic measurements outdoors, with the speakers mounted on a 50" turntable so that the speaker is as far away from any reflecting surfaces as possible, yet still manageable. I try to get 10 milliseconds of undisturbed sound between the initial impulse response and the arrival of the first reflection, which will give me a frequency resolution of 100Hz and useful data for all frequencies above a couple of hundred Hertz. I also try to minimize the first reflection off the floor or ground with acoustic absorbers. But as you can see, this method really doesn't tell you much about the bass.

After my series of anechoic tests, I perform in-room measurements over a 50ms time window. This gives me a frequency resolution of 20Hz, and since 50ms is a pretty long time in a room, it does take into account the room reflections. I also use 50ms because that is about the maximum time span [during which] the human brain can process the characteristics of a sonic event. Basically, I use these in-room measurements as confirmation of the anechoic results, not to correct for all the reflection anomalies or peaks and dips that show up in the response. I do, however, make these in-room tests from several different locations, and with our new dipole designs, even these in-room measurements over a long time window are surprisingly smooth and flat.

Another test I perform looks for resonances and stored energy in various locations -- using a Shaped Tone Burst stimulus, which is particularly well-suited for this. This is a tremendous test signal. I measure the impedance curve of the drivers themselves to reveal driver anomalies, and I also use complex multi-tone signals to test for nonlinear intermodulation distortion artifacts.

Dickson: This is essentially the spectral contamination distortion measurement you are speaking of?

Linkwitz: Yes, exactly, the same concept, in order to find nonlinear problems. Interestingly, in the old days when we used pink noise as the stimulus to try to equalize a speaker to be flat at the listening position in a real room, it typically turned out too bright-sounding. This is an approach that may be useful in a PA setup, but in a listening room it doesn't lead to a correct result.

Dickson: Tell us more about the Shaped Tone Burst test you just referred to. I found your article in the April 1980 issue of the JAES (Vol.28 No.2), discussing the benefits of using this stimulus in speaker evaluation, very interesting.

Linkwitz: From a practical standpoint, the advantage of using a shaped tone burst (one that rises and decays gradually in a sinusoidal envelope) is that all of the burst energy is concentrated into a very narrow frequency band. This is quite different from tone bursts used in the past, where you had a rectangular burst covering a fairly wide frequency band. I chose a spectrum width of a third of an octave for this stimulus -- which is a 5-cycle burst -- because this corresponds closely to how we hear. A third-octave is about the width of the critical band of hearing. Also, because the burst is so short in duration, you mask out the effect of reflections, so it becomes a sort of poor man's approach to anechoic measurements. As long as you measure the peak of the burst before the first reflection, you've essentially captured an anechoic-like response giving you some of the benefits of Time Delay Spectrometry or Maximum Length Sequence (MLSSA) techniques without the expense.

Now, the shaped tone burst can be used in several ways. For instance, one can just use a microphone to measure the peak amplitude that the burst reaches after you apply it to a speaker, which will give you an approximation of the frequency response. Likewise, after the decay of the 5-cycle burst, there shouldn't be any output from the speaker. In reality, however, if there is stored energy in the drivers or cabinet, the speaker keeps on ringing. Therefore, the shaped tone burst is very useful for identifying the sources of resonant storage. In any event, I do get extremely good correlation between the frequency-response measurements derived from the shaped tone-burst test and what we hear, as well as specific information about cabinet and driver resonances.

The real benefit of this type of test is that it concentrates the energy into a constant narrow frequency band so that it is a third-octave in width at 100Hz or 1kHz or 10kHz. Therefore, it is much narrower on an absolute basis at 100Hz than at, say, 10kHz. In other words, the tone-burst test has a constant resolution on an octave basis. This is important when you compare it to FFT analysis, where you get good resolution at high frequencies but very little information at low frequencies. The shaped tone-burst test works on a logarithmic scale so we can get good resolution all the way down to the lowest frequencies. I use the type of test signal to look at the decay of the burst, which gives me the same type of information that you would be looking for in a spectral-decay or waterfall plot that MLSSA can generate.

I also have MLSSA, so I do generate the spectral-decay plots as well, but I have to say, I have not found the waterfall plots very useful except for maybe above 1kHz. Below 1kHz there are so many artifacts in the typical spectral-decay waterfall plot that it is useless. Anyway, it's simply a lot easier to get the same, and even much more, information out of the shaped tone-burst response. Extending the time record for the FFT in order to get useful low-frequency data is generally not practical; using a narrow burst signal makes it so direct and easy. Plus, you change the frequency of the tone burst on the fly, while you watch the dynamic changes on a oscilloscope, as the tail of the burst stretches out -- in effect allowing you to see directly when you're close to a resonance!

I guess I'm beginning to sound a little like a missionary for the shaped tone-burst test, but I really do believe it is an extremely powerful technique that is too infrequently employed. Many people are just not aware of how it differs from traditional tone-burst stimuli. Today it is particularly easy to generate the required burst signals since you can buy an arbitrary waveform generator fairly inexpensively. Also, it would be very easy to include a series of 5-cycle-wide bursts at various frequencies on a test CD; then, with an oscilloscope or perhaps one of the PC-based softward test systems, the audiophile would be equipped with a powerful tool for evaluating his system and speakers.

One final attribute of the shaped tone burst that I find very important is that it's a particularly safe signal with which to test the maximum output of components. For instance, if you use a burst rate of 1Hz with a 5-cycle burst you'll have a very low duty-cycle, so even if you require 100 watts to clip your tweeter, the short duration of the burst -- it's essentially like a frequency specific pulse -- will prevent you from overheating the voice-coil and damaging the driver.

Dickson: You're most widely known as the developer of the Linkwitz-Riley crossover. Could you explain a few of the characteristics of this crossover?

Linkwitz: To answer your question, we need to go back to when I started out exploring the whole speaker issue in the early '70s. Then you could take the grille-clothe off many of the available speakers and see a strange, almost haphazard arrangement of the drivers on the baffle. It really puzzled me and I wondered what was going on. So I asked some of the designers why they were doing this and they said, "Because we've found it sounds better."

As I looked further into this issue, I realized that two principal things were not well-understood. First, very little was known at that time about the effects of diffraction from the cabinet edges. Second, and more importantly, very little was understood about how phase-shift with respect to the current passing through the voice-coils of different drivers affected the polar radiation pattern of a speaker. In other words, the interaction between the electrical side of a driver and the acoustical response was not clear at the time. For example, the phase-shift between the current in the tweeter and midrange voice-coils, relative to the placement of these drivers on the baffle, affects the speaker's radiation pattern.

Basically, since few drivers are really coaxial, with the difference in physical placement -- that is, if the path lengths between the drivers and the listening point are different, or even if they are the same -- you get a vector addition which is a function of the phase-shift between the different voice-coil currents and the distance between each driver and the listener. So Russ Riley and I began our work, in earnest, to be sure that the drivers were in-phase in the crossover region. This, in essence, is what the Linkwitz-Riley crossover is all about; making sure that you have the same acoustic phase between the midrange/woofer and the tweeter at the crossover.

Dickson: How about the phase relationship outside of the crossover region?

Linkwitz: As it turns out, that same phase relationship is maintained at other frequencies as well. This is very much in contrast to the classical Butterworth crossovers that people use in a number of speakers. An inherent property of the Butterworth design, whether these are first-order, third-order, fifth-order, etc., is that the crossovers are always in phase quadrature. In other words, the acoustical signals coming from the midrange and tweeter are phase-shifted by 90^o relative to each other. At its -3dB point, each driver has an amplitude of 0.7, and if you add two 90^o phase-shifted vectors of 0.7, you get unity -- the outputs of the two drivers add to unity on-axis. However, as you move farther away off-axis, one or the other driver will experience more phase-shift as the path-length difference becomes longer, and you'll have either a dip or a peak in the amplitude response off-axis.

In any event, the true maximum output of the two drivers will occur someplace off-axis, and this is an audibly bad thing. The peak off-axis response can then reflect from the nearest boundary and combine with the direct sound as added coloration.

Now, a first-order crossover can be made phase-perfect at one point in space, but I feel quite strongly that you cannot just look at a speaker's performance at one single point in space. The off-axis response is also very important to a speaker's overall performance in a real room, because the radiation in these other directions will add, through reflected and reverberant interactions, to what you hear. Typically, we don't listen to speakers outdoors or in anechoic chambers.

For an ideal Linkwitz-Riley crossover, the amplitude is flat on-axis or at unity, just as it would be for an ideal Butterworth. However, the Butterworth response will have its peak off-axis. In contrast, the amplitude of the L-R crossover will be down in level off-axis, and will never be higher than the on-axis response. The crossover point of a Linkwitz-Riley will also be at the -6dB point, equivalent to an amplitude of 0.5, and only when you add vectors with amplitudes of 0.5 that are in-phase will you get unity. If there is any phase angle between these half-amplitude vectors, their sum will be less than unity.

A very important point that people sometimes miss in this discussion is that when we are speaking of a given crossover, we are talking about an acoustic crossover, or what happens acoustically. Now, what I have to do electrically to achieve the correct acoustic response may not look anything at all like a textbook filter design. The actual filter often looks very little like the drawings I may show to explain any given example. This is also true for a Butterworth filter. It is highly unlikely that a textbook electrical Butterworth crossover will produce an acoustic Butterworth response, because the driver's response enters into the picture as well.

Dickson: There is a general misconception in some circles about differential vs absolute phase effects in speakers. Recently, I've heard about some well-meaning but misinformed retailers who arbitrarily reverse the polarity of either the tweeter or midrange hookup wires in all of the speakers they sell that are designed with high-order crossovers, in an attempt to make them "in-phase" -- much to the horror of the original designer. Perhaps you could shed some light on this issue.

Linkwitz: If someone were to arbitrarily change the polarity between drivers in a good Linkwitz-Riley crossover, they should get a strong null at the crossover point on-axis. In fact, this is a test I use to see how well I have executed the acoustic crossover. However, making such a change with the idea of somehow making a "phase-coherent" speaker is not correct. It will certainly change the sound, mind you, but is definitely not recommended.

Dickson: The Dvorak and Vivaldi speakers represent a radical departure from your earlier philosophy. What inspired this change in direction, and could you outline some of the primary goals you've tried to achieve with these new dynamic dipole designs?

Linkwitz: I would have to say the departure in my thinking happened by coincidence. At the time, I had volunteered to build a public address system to improve speech and sound intelligibility for a video production in a large, highly reverberant gymnasium. I designed a long directional column speaker with multiple 6" dynamic drivers firing as dipoles. In other words, the back of the column's baffle was open so the sound radiated to the front and rear with each direction out of phase with the other. The directivity of the radiation pattern of this design worked really well in this very reverberant environment. You could understand what was said just as well from the back of the hall as from the front. Well, just for kicks, I took the thing home, split this long column into two shorter stereo columns, and decided to see how it sounded in my living room. I was really surprised to hear that it had some of the qualities that impressed me so much with other dipole speakers, like the Quad electrostatics. Not that these PA speakers were anywhere close in tonal balance or transparency, but there was something fundamental about the character and quality that reminded me of these better dipoles.

In any event, that got me started on investigating the possibilities of using conventional high-quality dynamic drivers as dipoles. Quite frankly, I've always been very fond of certain characteristics that some electrostatic dipoles posses, yet I had never seriously pursued panel dipoles, in spite of their good qualities, because there was not enough dynamic low-frequency output for me. They also had a very tight listening spot, and were generally just too limited in what they could do.

Based on my initial experience with the PA columns, I set out to build a speaker that could perform similar to an electrostatic dipole but using conventional dynamic drivers in a manner that would avoid some of the limitations faced by large panel designs. It took several evolutionary iterations of the design to get a good understanding of which aspects of dipole operation are really important and which are not so critical, and why this is so.

During this development phase, I discovered how important the baffle shape and design were, as well as the frequency range over which dipole radiation was most beneficial. For instance, I found that dipole radiation of the treble region was not only unnecessary but, in fact, a disadvantage. Interestingly, you don't even see the rear-firing response from a dipole tweeter when you measure it on-axis, but when I listened to it in the room, I found that it caused some high-frequency "splatter" that didn't seem natural. I abandoned that approach and used a monopole dome tweeter from 2kHz on up and dipole radiation from the cone drivers down to 20Hz. I started out with a completely active system, with separate amps for all drivers, and used equalization as well. I equalized the speaker to be flat since a dipole on a small baffle has this natural 6dB/octave rolloff below a certain point. This first version was a pretty elaborate prototype.

Actually, at first I used a closed-box woofer because I didn't think I could get enough bass output from a compact dipole woofer system with a reasonable number of drivers. This is because of the acoustic short-circuit between front and back that a dipole represents. However, after sharing notes with Brian Elliott, a good friend and acoustician who had developed a splendid dipole bass system using six 12" drivers per channel, I began looking at the possibilities for dipole bass more closely. His system simply produced the most astoundingly natural low-frequency reproduction I had ever heard. However, I still stayed with the closed-box woofer concept a while longer. Then Don Barringer, another good friend and the recording engineer for the US Marine Band, reported great results in a normal-sized listening room using a speaker based on my design, but extending the dipole operation into the bass with two 12" drivers per channel equalized flat.

So, while I was refining the dipole midrange/monopole tweeter parameters of my speakers, I was actually a latecomer in taking the dipole concept all the way. As it turns out, and somewhat surprisingly to me, two big dipole woofers per channel does a very respectable job. Not quite like Brian's system with its total of twelve 12" drivers, but actually very good. On the other hand, we've also learned that there is very little difference between four and six 12" woofers per side, and with some creative mounting techniques, we've been able to mount four large drivers in a surprisingly compact dipole enclosure for use in larger rooms, and as a standard feature in our new top-of-the-line speaker system, the Beethoven.

Anyway, the production Dvorak is a five-piece, actively bi-amplified system comprising two main panels, two separate subwoofer cabinets, and an active crossover/equalizer. It covers the full audible range and beyond, from 10Hz to 25kHz. If a person has a small room or doesn't require deep bass below 40Hz, they can just use the main panel with a single amplifier, but the active crossover/equalizer is still required for the EQ of the two mid/bass drivers. We have also recently released the Vivaldi speaker, which has the same driver complement as the full Dvorak system but all mounted in one tall speaker. The system uses a passive design for both the crossover and equalization so it has a reasonable flat response to about 40Hz, below which the dipole's natural rolloff occurs. This system is ideal for those in small- to moderate-sized rooms who don't need the full 20Hz extension of the complete Dvorak system. It also requires just a single stereo amplifier.

Dickson: I suppose the directivity of the Dvoraks, particularly with the separate dipole subwoofers, presents some special considerations when these speakers are measured.

Linkwitz: One of the real problems was: How do you measure a dynamic dipole accurately? For instance, you cannot measure close to the cone because then you'll only see what the cone does essentially. You will not get the effect of the cancellation when the negative-polarity rear wave combines with the forward positive wave, so you have to measure some distance away. That really forces you into either an outdoor or anechoic-type measurement. It's very important to do it this way to get meaningful results. If the cabinet is large, or if the drivers are spaced very far apart, you have to have a least the same distance to your measurement microphone in order to capture the integrated sound coming from the rear, or off the edges of the cabinet, etc. The separate subwoofers are quite tricky to measure as well.

The bottom line on the Dvorak is that everything is based on a flat response under anechoic conditions with a moderately directional radiation throughout its full range. When I say that the dipole aspects of the speakers are directional, it's not in a very strong sense. For instance, the response is 1dB down at 30^o off-axis, 3dB down 45^o off-axis, 6dB down at 60^o, and has a null around 90^o. The monopole tweeter also maintains a similar directional characteristic because of the baffle design and the wavelengths of its frequency band. Of course, it differs from the dipole drivers in that it fires predominantly forward. The important thing is that the shape of the off-axis room response is very consistent with that produced on-axis, resulting in the open-sounding soundstage and the speaker's even tonal characteristics. Also the dipole "figure-8" cosine directionality goes all the way from 2kHz down through the woofer's range to 20Hz. This directional deep bass is really pretty amazing! If you have another person to help you, play some low-frequency tones, then have your friend rotate the woofer cabinet and you can clearly hear the output null at 90^o off-axis!

Dickson: Yes, I've noticed I can hear increased focus in the low bass when I toe-in the woofers. This is one speaker where "stereo" bass may have some real meaning. Two of the Dvorak's sonic characteristics I find most striking are its dramatic reduction of room-induced colorations from the low bass through the midrange, and its ability to convey image height in correct proportion to the width and depth dimension of the soundfield. What factors do you think contribute most to the effects?

Linkwitz: These are primarily due to the dipole characteristics and the even room response. Since the speaker is moderately directional at all frequencies, more of the sound is directed toward the listener and less to the walls and ceiling. Therefore, less comes back from the room in the form of resonances or reverberation which will blend with and color the direct sound from the speaker.... The active equalization is merely there to correct for the dipole cancellation that the raw drivers would have if you didn't compensate for the inherent 6dB/octave rolloff.

Now the image height is an interesting thing. I have to state that I don't fully understand all the psychoacoustics involved here, but I have found that it is important that the center of the radiating elements be at about ear height, and that the speakers have some vertical extension as well. I have built many small two-way minimonitors; while these systems can have very nice horizontal dispersion and excellent imaging, I've always felt that I was listening through a horizontal window, one that was very wide but with a height not much greater than that of the speaker itself. It's like listening through a horizontal sliver. Now vertically spreading out the driver's arrangement expands the vertical dimension of the soundstage and adds much more realism for me.

With respect to the reduction in overall room colorations that the Dvoraks provide, that has a lot to do with how the dipole characteristics are implemented. This comes back to the fact that the off-axis response is very well-behaved in this system. In other words, the design concentrates just as much on the off-axis performance as on the on-axis. While I don't have any definite proof, I strongly suspect that the erratic off-axis behavior of most panel speakers is what makes their room placement so critical, forcing a person to locate the panels in a place that minimizes reflections and changes how the off-axis sound couples with the room, in order to get a balanced output. On the other hand, quite frankly, I have not found the performance of either the Dvorak or the Vivaldi to be critically dependent on room placement compared to other speakers. There is still definitely an optimal placement in any given room, but you can get very satisfactory performance in a wide variety of locations, so this experience lends further credence to the value of a well-behaved off-axis response.

Dickson: While the Dvorak and Vivaldi represent a somewhat fresh approach to speaker design, they are mature designs. I'm very curious to hear about what projects you have planned for the near and more distant future.

Linkwitz: Recently, I've been doing extensive investigations into numerous drivers using some of the newly developed measurement techniques I alluded to earlier -- especially the tests for nonlinear distortion artifacts, and those that help locate and define energy-storage effects in drivers themselves. All this in a search for components that have even more clarity and transparency. What I had in mind was to see how much further this Dvorak concept could be refined.

We unveiled our new flagship, the Beethoven, at the recently completed '96 WCES in Las Vegas. In addition to an all-new balanced electronic crossover, each main panel has a new silk-dome tweeter, two new 8" drivers, and a pair of 10" dipole drivers -- all low-distortion, high-excursion models. Both of the woofer cabinets for the new system contain four 12" dipole drivers, so obviously this system is designed for high-output, very-low-distortion sound and will be considerably more expensive than the standard Dvorak. I must say we have been extremely gratified with the performance of the new system. So that's one project we are putting the finishing touches on now, and we are also thinking about a smaller, lower-cost version of a dipole speaker in the future.

A little farther down the road, possibly over the next few years, I would also like to settle in my mind the importance of what I would describe as "linear phase." This refers to obtaining a result that is a more accurate replica of the time-do-main wavefront. Some people seem to think that this is very important for reproducing clicks and transient-type sounds, and that may well be. From a common-sense point of view, it seems logical that you would want to have a true replication of the wavefront.

However, I'm not totally convinced because I have done a lot of experiments with phase-distorted signals. Basically, I've shifted the phase between different spectral components by running various signals through an "all-pass" filter, where the amplitude is unaffected when I change the phase response with frequency. When you look at these signals on a oscilloscope and change the phase, they look grossly different, so you'd think "surely this must sound different." But when you listen, you can't hear the difference, even though the time-domain waveform staring you in the eye looks so totally altered!

I did quite an investigation into this when I initially developed the Linkwitz-Riley crossover because it is not a linear-phase system -- nor are the Butterworth crossovers, for that matter, except for the first-order slope. The experiments I've done so far have not convinced me that phase distortion in small amounts is audible. Now if the phase distortion is gross, you can definitely hear it, but the typical crossover is far from producing that much phase distortion. However, some people whom I respect seem to think this is something that could have audible consequences, so I'm keeping an open mind about it and want to determine once and for all its value, if I can. I must say that I have not heard an example of a speaker design that conclusively demonstrates the benefit of a linear-phase system.

Dickson: I imagine when you look at the total performance of a speaker with the various tradeoffs required to achieve a certain goal, you have to weigh their relative merits.

Linkwitz: This is true. You could question, for example, whether the extra stress on drivers and resulting distortions produced by a first-order system are not more audibly significant than the subtle improvements potentially created by its linear phase effects. However, it is possible, using digital techniques, to correct for the phase response as well, and my friend, Malcolm Omar Hawksford [of England's Essex University], who has done quite a bit of work in this area, has kindly offered to perform a phase correction for the first 10ms time record of the Dvorak's impulse response with his digital processor. It certainly would require a bit of horsepower to implement digital phase correction in the active crossover, but it could be done.

Again, it's not yet completely clear whether a digital crossover will buy you anything. It may buy something, and that's the part I'm interested in. For instance, with this scenario we could combine the excellent on- and off-axis amplitude response of the existing Linkwitz-Riley crossover in the Dvorak with an after-the-fact digital correction of the time-domain, to achieve a linear-phase system. You see, the digital time-domain correction would not affect the existing passive or active crossover response at all, it just would correct overall phase. As a matter of fact, at a recent AES convention, both Malcolm Hawksford and I attended a discussion about the use of a very steep crossover filters with digital phase correction. Convincing arguments were presented showing that these extremely steep filters produce sonic anomalies, and consequently are not desirable. Malcolm also stated that something like the Linkwitz-Riley fourth-order crossover was about optimum, even digitally implemented, when phase correction is applied. Anyway, I'm very interested to see how this research turns out.

Dickson: With the continual improvement in driver technology and refinements in other area of audio design, it may be that these more subtle issues, like linear phase, will become more important in the future.

Linkwitz: I think that's a good way to look at it. You could say that you need to have a certain number of other things done correctly first before those effects come into play. I should also point out that the digital phase compensation I'm speaking of is very different from the digital room-correction systems you may have read about. In any event, these are a few of the areas that we at Audio Artistry look forward to investigating and developing in the near future.