There are two different types of people in the world, those who want to know, and those who want to believe.” – Friedrich Wilhelm Nietzsche
Alon Wolf writes: “As a service to readers who, in Nietzsche’s terms, “want to know,” we will devote a series of columns to the science and engineering behind Magico design choices. A quick glance at online discussions, some manufacturers’ marketing material and even product reviews reveals that key principles are not widely recognized, much less understood. We aim to use plain English to articulate the fundamentals of loudspeaker design and how those fundamentals shape our engineering efforts. In the process, we’ll discuss the difference proper design makes toward uncompromised music reproduction.”
At Magico, our number one priority is to minimize the losses in electro-acoustic transfer, the conversion of recorded music into acoustic energy. We aim to preserve maximum fidelity to the source. We’re not out to beautify or manipulate the recorded signal. We’re not looking to add performance attributes or exaggerations, whether on purpose or through lack of knowledge. An exaggeration or attribute not in the original recording may catch the ear at first, but lead to long-term fatigue, endless system tweaks and ultimate dissatisfaction.
Proper speaker design empowers reproduced music to take us into the same “blissful zone” as live music. Backed by decades in loudspeaker design, we have identified the single most important factor in achieving that blissful zone.
It’s equilibrium. Equilibrium demands that no single aspect of performance can outweigh the others. This may sound easy, but it imposes serious constraints. For example, if a designer chooses brilliant highs with AMTs or ribbon tweeters, the tweeter response can’t blend seamlessly with a point source woofer of completely different acoustic behavior. And while we appreciate the dynamic range of horns, there’s no way the limited bandpass of a horn midrange can blend seamlessly with a conventional woofer. It just can’t happen, no matter how hard a designer tries or what spin a marketer puts on it. As we’ll see, these issues of driver selection are not isolated examples. Techniques that excel in only one acoustic region will penalize the design with unsolvable deficiencies.
TOPIC 1: Equilibrium and driver placement
The quest for equilibrium encompasses every aspect of loudspeaker design and manufacturing. The required disciplines include materials science and acoustics in addition to electrical and mechanical engineering. Designers need to understand them all as they apply to loudspeaker performance.
A prime example of how these disciplines interrelate is the first subject in our series, driver placement. The importance of proper driver placement becomes clear when you consider the scientific research. In blind experiments, listeners expressed one consistent preference. The speakers rated as highest in sound quality were the ones with the smoothest directivity plot (similar to, but not the same as, a smooth dispersion pattern). The smoother the plot, the better a speaker sounded.
Although many believe that “good” or “bad” frequency response refers only to on-axis performance, the reality is much more complex. On-axis response is only a small piece of the puzzle (and concentrating on attributes like flat on-axis response can lead a speaker designer into worlds of trouble). When a loudspeaker’s individual drivers release different types of dispersion patterns into 3D space, they generate summations and cancellations that are radically different from the recorded signal. At a subconscious level, your brain feels that something isn’t right, an irritation that prevents you from entering that “blissful zone.”
To understand why a smooth dispersion pattern is so critical, we need to appreciate some psychoacoustics, the science of how the human brain interprets sound. Pioneering research conducted in the late 1940s and early 1950s discovered that the brain interprets two identical sounds presented in close succession as a single, fused sound. This is called the Haas Effect. For clicks, fusion occurs when the lag is less than 5 milliseconds. But for complex sounds like speech or piano music, fusion occurs when the lag is less than 40 ms. When the lag is longer than 40 ms, we hear the second sound as an echo. This 40 ms threshold equates to 45 feet (14 m) of travel at the speed of sound. As a result, to the human brain, early room reflections in typical homes will be indistinguishable from direct speaker sound.
There’s more. The vast majority – as much as 80% – of the sound you hear in a typical listening room reflects off walls, celling and floor. The on-axis frequency response describes as little as 20% of what you hear. It’s critical that the delayed signal (off axis) be as similar as possible to the original signal (on axis). That is how we hear the world around us. Sound sources like voices or musical instruments generate sound in a smooth directivity pattern.
Room reflections account for as much as 80% of the sound we hear.
And the first reflections are typically indistinguishable from the direct speaker sound.
Setting smooth directivity as the top priority really limits the ways engineers can design and locate woofers, midrange drivers and tweeters.
All drivers must have a smooth directivity plot not only in their individual frequency bands but also in their complex sums. The latter requires meticulous crossover design.
Drivers must align vertically, facing the front, so the waveform will propagate from the same plane.
Due to their increased directivity, tweeters must be at ear level. Tweeters must also avoid sharp enclosure edges, which trigger diffraction and its consequent interference patterns.
Above 600 Hz, there must only be one driver per bandpass. While two drivers can boost efficiency, using two or more drivers degrades coherency and muddies the imaging. Discontinuities in the polar responses due to the physical separation of the drivers will result in phase differences leading to destructive interference in the driver’s passband at various angles.
Firmly based on physics and psychoacoustics, these principles are the foundation of every Magico loudspeaker design. They dictate so much of what we do, including how we choose driver efficiency, driver size and cabinet shapes. They even motivate our choice of Elliptical Symmetry Crossovers with 24 dB per octave slopes.
These principles can be quite restrictive. For example, it would seem entirely benign to move the tweeter from ear level and simply tilt it toward the listening position, in violation of principles #3 and #2. But this actually degrades off-axis response.
We can see the results graphically. Taking a well-designed speaker and placing the tweeter at ear height results in the smoothest distribution, as shown in the first chart. Here, the polar plot shows the amplitude of frequencies near the midrange/tweeter crossover in decibels from +90° (straight up) through 0° (listening height, approx. 40 inches/102 cm) to -90° (straight down). The curves are well controlled.
On the left, placing the tweeter at ear height produces the smoothest dispersion pattern.
On the right, changing the speaker location and adjusting the angle makes the dispersion go haywire.
Next, we take the same speaker, change the location and adjust the angle. As one might expect, the 0° response is almost identical. But the off-axis response is substantially disrupted. Designers can try adjusting the crossover or driver geometry, but the fundamental problem remains. The room reflections will vary dramatically. Musical coherence is lost. Soundstage is degraded. Imaging suffers. What seemed at first to be such a good idea turns out to be a problem.
This simple change in driver geometry is not the only example of design choices that may seem benign but lead to eventual disappointment. Many other topologies excel in one specific area, but create problems elsewhere.
- In an attempt to achieve time alignment, setting individual drivers back from the listener creates physical steps in the front of the speaker. This will lead to diffraction, which according to listening tests is far worse than the unproven benefits of any physical alignment. True time coherence is itself an elusive goal. Crossovers introduce significant, frequency dependent delays. Minimizing those delays leads to big compromises in nonlinear distortion. Even impulse-response waveforms that purport to demonstrate time alignment at one listening position say little or nothing about overall speaker performance.
- D’Appolito or MTM driver configurations have great sensitivity and make for easy driver placement. But they suffer from unsmooth directivity that additional engineering just can’t fix.
- Line arrays have low distortion and great dynamics. But their very narrow directivity severely restricts placement.
- Side-firing woofers can make for narrow, attractive cabinets. But a low cutoff frequency results in a very inefficient crossover – negating a key benefit of large drivers: reducing intermodulation distortion in the midrange.
As in so many areas of life, when it comes to loudspeaker design, you can’t get something for nothing. Everything has a price. If you count yourself among Nietzsche’s “people who want to believe,” you may be willing to overlook these performance penalties. But if you “want to know,” you’ll approach loudspeaker choices with your eyes wide open.
1. Much of the loudspeaker marketing behind time coherence has been misleading. To begin with, there’s never been any scientific proof that time/phase coherence improves perceived sound quality. In addition, true coherence requires meeting two conditions: 1) a first-order acoustical crossover, that is, a perfect 6 dB-per-octave slope from the designated bandpass; and 2) the physical alignment of the drivers’ acoustical centers, which, unless a concentric driver is used, only achieves coherence at one sweet spot in 3D space. Just moving drivers around is not enough. In fact, such designs ensure performance losses. If a first-order crossover is not used, any driver movement will require crossover realignment to keep the proper phase relations among drivers at the crossover points.
There have been honest attempts at such designs, including some that do meet the basic conditions. However, even when designs meet these criteria, sound quality suffers.
- A 6 dB-per-octave slope in a typical three-way design means that the bass drivers will be only about 18 dB down at 2 kHz – playing right into tweeter territory. And the tweeter will play into the bass region. This introduces big, easily audible increases in intermodulation, second harmonic and third harmonic distortion. An accurate 6 dB-per-octave crossover is also very complex, with many parts that themselves degrade sound quality and introduce time delays. A simpler crossover using non-pistonic drivers is possible. But non-pistonic cone movement increases distortion and destroys musical nuances.
- Staggering drivers, in order to align them in a stepped baffle creates tremendous diffraction. While the supposed benefits of time coherence are unproven, the audible distortions of diffraction are very real.
The unavoidable tradeoffs of even a “successful” time- and phase-aligned design also include tilted drivers firing at the listening position asymmetrically, limited vertical dispersion and reduced power handling. After weighing all these tradeoffs against the lack of evidence that alignment actually contributes to sound quality, our decision to reject time- and phase-alignment was easy.