Now it’s time to get down to the nuts and bolts of how speaker drivers really operate, and where they depart from the ideal. Despite the physical simplicity of a classic Rice & Kellog direct-radiator, the conversion from electricity to sound has three steps:
1) Electric power to magnetic force - the voice coil
2) Magnetic force to mechanical acceleration - the magnet and pole piece
3) Mechanical acceleration to acoustic radiation, via direct radiation and/or augmented by an enclosure or a waveguide (horn).
There are characteristic departures from the ideal at every step of the way, which we’ll be looking at in the following paragraphs. This is a good point to draw your attention to marketing claims of “removing” this or that coloration - as you’ll see, complete removal is impossible (contradicted by physics), and techniques to reduce problems in one area quite often make problems in another area much worse.
There are other ways to accomplish the electric-to-acoustic transformation - electrostats, piezoelectric transducers, or even ionized and modulated air - but these have their own set of problems, which are even more intractable than the problems of the conventional direct-radiator.
1) Electric power to Magnetic force. This looks simple - it’s nothing more than a voice coil wound on a non-magnetic former. Nothing more than a small air-core inductor, right?
Copper, aluminum, and silver (the materials used to wind voice-coils) all have temperature coefficients. With the voice-coils used in typical 87 to 91dB/metre audiophile speakers, the increase in DC resistance due to local heating effects results in 1dB of compression at levels as low as 90 to 95dB (average). If it were instantaneous compression like a good broadcast compressor, that wouldn’t be so bad - but it isn’t. The voice coil heats and cools with a time constant of several seconds - the time it takes to transfer its heat into the magnet assembly. In addition, in a multiway system, the time constants for the woofer, midrange, and tweeter are quite different, so the compression effect not only fails to track the music, but lags behind it with different time constants in different parts of the spectrum.
This is one of the primary benefits of a high-efficiency system - for a given sound level, the voice-coil doesn’t get as hot as a lower-efficiency system, so there is less compression due to VC heating. This is a direct benefit of high-efficiency prosound gear - in the home, it is operating at levels 10 to 20dB below professional use. The tradeoff - and there’s always a tradeoff, folks - is degraded performance in the time domain, and rougher response in the polar pattern and in the frequency response.
There’s an odd fetish out there for 16-ohm speakers - which is good for transistor or OTL amps, but doesn’t make a lot of difference when an amplifier is transformer coupled (the transformer already has a 20:1 stepdown ratio at the 16-ohm tap; 40:1 isn’t much different). One area where high-impedance voice-coils are at a serious disadvantage is higher mass; the mass of a voice-coil is parasitic, playing no role in sound radiation, and degrades the HF response fairly significantly, in addition to inviting in additional HF resonances.
How significant? Well, in a widerange driver, it can mean the difference between smooth response out to 8 kHz versus somewhat rougher response out to 5 kHz. In all honesty, if an output transformer can’t deliver full performance at 8 or 4 ohms, maybe you should look elsewhere.
2) Magnetic force to mechanical acceleration.
This is where we look at magnets, and what they’re doing to that little voice coil. On its own, the magnetic forces generated by the VC assembly in free air are tiny; the pole-piece and magnet assembly multiply it in the same way an iron-core multiplies the inductance of an air-core inductor. In addition, the magnet supplies the reaction force which the multiplied VC pushes against, generating movement.
Now we’re getting into deeper waters. As we know from transformers and iron-core inductors, the choice of core material makes a significant difference to the sonics, since magnetic systems are not linear (unlike air). The relatively small value of inductance of the VC in air is greatly multiplied by the magnetic system of the loudspeaker - and this new inductance is always in series with the circuit.
In bass, mid, and full-range drivers the inductance starts to become significant above 500 Hz. To compensate for self-inductance, most drivers are mechanically peaked, or exhibit a rising response, above this frequency. This is why a driver that should be rolling off above 500 Hz to 1 kHz, will actually have flat or rising response out to 5kHz.
As with many moving-magnet cartridges, the mechanical system balances out the electrical system. This also why simply adding inductance to a woofer crossover results in an annoying shelf response, instead of the expected rolloff. If the woofer had mechanically flat response and no inductance, adding an inductor would give the expected 6dB/octave rolloff. But since it’s already carefully balanced mechanically and electrically, all doubling or tripling the self-inductance does is create a more complex response, typically with a shelf characteristic.
Inductance compensation, or a Zobel network, isolates the crossover inductance from the woofer inductance. A more sophisticated approach is to use a computer-based “crossover optimizer”, but this requires precise measurements of the acoustic and electrical response before it can optimize a textbook crossover to one that really works.
The magnet is such an important part of the overall driver that one of the most important specs is the cryptic-sounding “BL Product”, usually in Tesla/Metres. BL product is actually really simple: it’s nothing more than the magnetic force in the gap (B) multiplied by the length of wire immersed in the magnetic field (L). In other words, this is the horsepower of the motor! The other key spec of the speaker is the total moving mass; these two, just like a car, control how fast it can accelerate. More BL, less mass, the faster it goes. Less BL, more mass, the slower it goes. In a loudspeaker, more acceleration translates to louder sound.
This is why a small gap is very important; the lines of force are more tightly concentrated. It’s also why large-area magnetic-planars are not very efficient; the BL product is much lower than a typical direct-radiator. The perfectly flat impedance curve is a direct reflection of the low BL product (loose coupling); if the BL product were higher, you’d start to see VC inductance, and if it was really high, cone resonances start to appear on the impedance curve.
The amplifier designer’s desire for a purely resistive speaker is a hopeless dream; the worst speakers are resistive, while the best ones, with the most powerful BL product, are highly reactive, and faithfully transmit all of their many resonances back to the amplifier. It is the responsibility of the amplifier design to do their job and design an amplifier that is not affected by reactive loads, instead of pushing it off onto the speaker designer to make speakers that are high-impedance and appear like resistors. If the amp designer can’t handle the real world, maybe they should stick to headphone amplifiers.
One mystery about magnets is why different materials sound different. These differences don’t show up on MLSSA in frequency response and waterfall charts, and don’t jump out on IM measurements, either. But different magnets do sound different, in a way that’s hard to pin down subjectively. It’s not a coloration in the usual cone-material kind of way, it sounds more like differences in the amplifier circuit or the choice of coupling caps vs transformer coupling.
As mentioned earlier, the voice-coil sends magnetic forces into the magnet assembly, and the magnet assembly reacts to this force in different ways depending on whether it’s made of ceramic, AlNiCo, Ticonal, Neodymium, or soft iron with a field coil to energize it. The choice of material affects the eddy currents induced by the voice coil; ceramic magnets are electrically insulating, while AlNiCo is conductive. If you were going to look for differences in IM distortion, 500 Hz and up would be the place to start, since this is where inductance starts to be significant.
And below 500 Hz? This is where the geometry of the gap becomes important, since the VC moves much more at lower frequencies (excursion increases at a rate of 12dB/octave for a direct-radiator). The linear-travel region of a tweeter or midrange is typically 1mm or less, so if you can see them moving, they’re distorting. The linear region of a hifi woofer isn’t much more - typically, 6 to 12mm. When you see claims of 25mm, most of the time, that’s excursion before damage, not the linear region.
There are a handful of specialized (and expensive!) audiophile sub-bass drivers that claim linear travel of 25mm or more. Before you get too excited, glance at the efficiency part of the specifications - it can be pretty low, like 85dB/metre. As mentioned above, low efficiency also means early onset of VC heating and compression effects. Once again, tradeoffs. I like to go in the other direction, efficient prosound drivers that are rated for hundreds of watts in continuous use, at least for bass drivers.
3) Mechanical acceleration to acoustic radiation.
This is where almost all driver coloration is coming from. There are two sets of independent problems: resonances in the spider, surround, dust cap, and cone, and antenna-radiation beaming due to cabinet diffraction and standing waves, lobing due to multiple drivers working at the same frequencies, and lobing due to cones being larger than wavelength they’re transmitting.
All of these are audible, although the antenna-radiation problems are not that important for single-speaker playback of mono. For mono, a big speaker with lots of drivers is actually an asset, since it can give a sort of pseudo-stereo effect thanks to its large area. Speakers with near-perfect point-source radiation can be annoying to listen to in mono, unless you play them in groups, which opens up the soundstage.
For 2 or more channels, though, radiation patterns are important, especially in reducing listening fatigue. Stereophonic sound, unlike old-fashioned stereoscopes, does not recreate reality. For sound, reality is a series of coherent wavefronts coming from all different directions. 2-speaker stereo is at best an approximation of this - even the best 2-speaker stereo has artifacts, such as comb-filtering and elevation of central sound sources (like a singer). This is the best-case; if the speakers have appreciable diffraction, multiple unsynchronized sound sources, or a complex polar pattern, then the near-3D picture will disintegrate into hard Left, hard Right, a wobbly, phasey Center, and a vague wash of sound filling the area between the speakers. Most audiophiles, and I suspect most reviewers, have never heard “real” stereo, even once.
When I was working at Audionics in the late Seventies, I built experimental speakers with almost no diffraction and good impulse response; in a dark room, you could walk right into them while they were playing, since they gave no impression of a loudspeaker at all. As a physical sound source, they completely disappeared. The soundstage, or rather, impression of acoustic space, had no apparent boundaries, was far larger than the physical dimensions of the listening room, and extended across 120 to 160 degrees - all of this with ordinary stereo recordings played on a good-quality phonograph.
Most people have never heard this; for them, it’s stereo if they can hear Left, Right, and Center, and get some kind of impression of dimensionality. When you pay attention to diffraction and the polar pattern, though, you get what I described above. That is the potential lying within most stereo recordings.
The dominant source of coloration, the quality that gives away the game that we’re listening to a mechanical contraption, are resonances in the cone, spider, dustcap, and surround, along with an assortment of cabinet resonances. Some audiophiles, especially who listen to high-efficiency systems, have trained themselves to “tune out” these colorations. This is a subjective matter; people really do hear things in different ways, and culture plays a role too. Americans, Brits, Germans, French, Italians, and Japanese have quite different national “styles”, especially in preferred loudspeaker design.
But there are some signal sources where ear-training, and culture, don’t make a difference. Pink-noise and applause immediately give away cone coloration; with these signal sources, the ear is 10 to 15dB more sensitive to coloration than it would be otherwise. Pink-noise should sound just like falling water, or the surf; the fact that it never does is a comment on far we have to go in speaker design. It’s still a good tool in refining driver and crossover design, though, since every time you remove a coloration you can hear it right away. The BBC started using pink-noise as a subjective reference tool back in the late Fifties, when they made the discovery that the ear can much more easily detect coloration with noise than recorded music.
It should be kept in mind that the BBC is one of the few organizations in the world where a speaker designer can do a direct comparison between a live orchestra and the prototype monitor speaker, simply by walking out of the control booth. For the rest of us, balancing a speaker with a few favorite recordings - made with unknown methods and equalization - and electronics - with their own set of colorations - is a pretty tricky thing to do. Although pink-noise is tedious to listen to, it has the advantage that the sound almost entirely unaffected by amplifier coloration. There’s also the further advantage that a speaker that is subjectively balanced to be flat on pink-noise almost always sounds well-balanced on music.
Returning to driver coloration, a primary source of reflections is the junction between the voice-coil and the cone, which are usually made of dissimilar materials - with a different speed of sound. If the dustcap is made of felt and placed right on top of this junction, it can damp the reflection and smooth out the response - unfortunately, dustcaps are usually only decorative, and worse, placed well outside the VC-cone junction, so it creates additional reflections. Some drivers use no dust cap at all, and have a bullet-shaped “phase plug” instead. My experience with these has not been positive - the metal bullet only seems to create additional reflections of its own - not what you want at the most sensitive location of the entire cone.
Exotic cone materials - Kevlar, woven carbon-fiber, fiberglass, aluminum, titanium, etc. have the advantage of greater rigidity at low frequencies, which decreases IM distortion, improves dynamics, and provides a greater sense of clarity. But as alway, no free lunch. These same materials, thanks to their greater rigidity, have more violent breakup modes, typically starting at 3 to 5 kHz, and extending all the way out to 20 kHz. As with horn enthusiasts, many audiophiles don’t hear the cone-breakup coloration. Some magazine reviewers actually confuse the breakup coloration for “resolution” or “detail”.
Uh, no. Musical instruments don’t usually have metallic tin-can colorations, nor do they sound like ripping paper, which is the characteristic sound of rigid materials when they break up - and don’t kid yourself, nearly all cones exhibit breakup behaviour. This is easily measured with MLS systems - in the 3D waterfall curves, look for regions above 3 kHz that look like mountain ranges - and also in the excess group delay vs frequency measurement, where you can see where the center of radiation is breaking into several incoherent groups.
A sharp enough crossover - 4th-order, for example - can usually suppress most of the coloration. But crossovers don’t work miracles - there’s usually a some subjective residue left, even if it’s been suppressed 20 to 30 dB. I prefer drivers that sound musical with as little equalization as possible. This is easily auditioned on the same IEC-style test baffle used for making measurements. An IEC baffle is nothing more than a plywood panel, with the driver mounted slightly off-center to minimize standing waves. In addition to giving much more accurate driver measurements than measuring in-cabinet, it’s also an outstanding way of doing quick evaluations of the sound of the driver itself. Remember, a cabinet is not going to improve the sound of the driver, only reinforce the bass range - and at the expense of midbass-region box colorations, in the 200 to 500 Hz region.
(An IEC test baffle for an 8-inch driver is 135 by 165 cm, for a 10-inch driver 169 by 206.5 cm, for 12-inch driver 202.5 by 247.5 cm, and for a 15-inch driver 253 by 309 cm. Click here for recommended driver locations on the panel. For auditioning, and most measurements, these dimensions are not critical.)
Well, we’ve covered some of the basics of driver design, and the most important sources of driver coloration. This leads to the next topic, what you want the crossover to do - reduce IM distortion (the most important function), reduce coloration by steeply rolling off or equalizing the frequency response, making the response function fit the desired curve, which in turn controls the polar pattern at the crossover frequency.