On Loudspeaker Directivity Part
you go any further, Part
1 is must-read (this article won't go very far without it). You
particularly need to look at the definitions and how to read the polar graphs
A dipole is, most simply, a loudspeaker system that sends sound equally to the front and rear, but with opposite polarity. A speaker without an enclosure is a good example -- when the cone moves forward, there's pressure applied to the air in front of the cone (compression), and suction applied to the rear (rarefaction). Pretty straightforward right? And it is -- if the pressure were a plunger in a tube. But our "plunger" is vibrating back and forth, with varying frequencies and intensity, and without a tube to contain the pressure or suction. These forces tend to meet at the sides, and cancel each other out. The more "tube" you use to contain it, the lower the frequency this cancellation happens at. That "tube" in the case of dipoles is a baffle. The baffle can be folded or manipulated any number of ways, and ultimately that is the hard part about designing a dipole -- the long wavelengths in the bass from the front and the back of the cone want to wrap around the baffle and cancel each other, where compression meets rarefaction. Conceptually, think of a pair of flashlights back to back. The light in a dark room tends to form a figure 8 in this arrangement, and this is a very rough approximation of how it behaves in a dipole loudspeaker.
When the baffle becomes small relative to the wavelengths, the front and rear begin canceling each other and less energy is produced. The larger the baffle the greater the pathlength between front and rear and the less cancellation happens. There is generally a peak, the "dipole peak" where the front and rear wave meet and are in-phase due to the different lengths between the front and rear. Below this peak, SPL rolls off at 6dB/octave until Fs (the resonant frequency of the driver) at which point it rolls off at approximately 18dB/octave (room involvement and Qts play a role too, however). You can see an approximation of how this rolloff works. There's a peak around 400 Hz, below which the SPL rolls off, and particularly severely off-axis.
To push the dipole peak frequency and beginning of this energy
cancellation down to lower frequencies, baffles are often very wide and
measuring at 48" or more. Alternatively, many people will use wings or other
open-ended assemblies to make the baffle appear wider. One such solution that's
clever in its simplicity is Nelson
Pass' "Slot Loaded Open Baffle"
Why would anyone want to use a dipole? Part of the advantage of dipoles is that the cancellation is actually desirable in some ways -- if sound isn't radiating to the sides because of this cancellation, it is not producing as many early reflections off the sidewalls of a listening room. There are reflections from the rear facing "lobe" but they are delayed and attenuated, traveling much further before encountering a reflective back-wall and coming back towards you. The ear is better at ignoring these delayed, lower SPL reflections than it is at ignoring the reflections coming from the sidewalls, which tend to be both louder and less delayed. Dipole reflections tend to be perceived as spaciousness, which is generally a pleasant characteristic. The problem here, is that the limitation of radiated energy to the sides due to cancellation) is maximized when the baffle is minimized- in other words, you need the most equalization and additional amp power and speaker displacement. If you're willing to go with a "nude" driver, it'll have fairly good constant directivity behavior below the dipole peak (which will be fairly high relative to the driver size). The larger the baffle gets, the less consistent the directivity behavior will be as the baffle will provide more interference to the directivity imposed by cancellation. An ideal, full bandwidth dipole can maintain excellent directivity behavior across its bandwidth.
It is important to keep in mind that Dipoles can suffer the same challenges of inconsistent reflected energy and power response nonlinearity of systems like the two-way 6.5" in part 1, the dipole action takes place largely from the midbass down. Constant directivity horns could be used in a dipole, though finding a good transducer/horn combo to create the dipole wavefront could be challenging. Of course, some of us like to get funky, and use ribbons, or AMTs, that are symmetrical front/back, and maximizing this symmetry is important for overall sonic quality. My dipole Heil AMT tweeter horns
My dipole speaker with monopole bass below 250 is comprised of the Heil di-pole and a pair of nearly -- baffleless high efficiency 6.5"s. A rough representation of the polar behavior is below -- as you can see, the directivity is fairly consistent from the lower midrange on up.
Well used, dipoles can create a very high performance soundfield. Many attribute their character to a lack of boxy coloration. Panels themselves can resonate, so even though they're not trying to contain pressure, they still must be properly braced. Simply being a "dipole" doesn't excuse you from getting the crossover, "cabinet" (such as it is), quality drivers, or any other of the things that a box speaker needs, it's just a different set of compromises.
Anyone remember what baffle step is? Baffle step is the action by which sound wraps around the baffle and radiates into full space rather than into the front hemisphere only. Bipoles deal with this quite elegantly, by introducing a second source that "fills in" the reduction in SPL caused by the woofer radiating into full (4pi) space rather than forward only (2pi). Accordingly, the on-axis response doesn't need the same adjustment or baffle step compensation. This is a significant advantage in maintaining system efficiency, but requires twice the drivers and thus twice the enclosure. No free lunch here!
Bipoles aren't very common in today's loudspeakers. One reason
for this is cost- they require twice the drivers, one set for the front, one for
the rear. Another is that like dipoles, they need a decent amount of distance
spaced out from the rear wall, which is a challenge in many rooms and thus
limits the available customer base for commercial efforts. Two manufacturers
making full-range bipoles are Definitive Technology and AudioKinesis. They are
very different approaches, with Definitive having front and rear midwoofers/tweeters
on a slender baffle. These are cone and dome types, and unfortunately likely
have much overlap with the "two-way 6.5" speaker" as discussed in part 1.
Remember that dispersion type, be it dipole, bipole, or monopole, doesn't
eliminate the need to take efforts for constant directivity, if you want a
balanced mix of room and axial responses.
Duke LeJeune of AudioKinesis took a different approach to bipoles, as shown here. The ideal version of his pattern is shown below. As you can see, at further off-axis angles, the SPL starts rising in the bass.
is a constant directivity waveguide two-way similar to the two-way from my
project below (but with two sets of them, front and back). Also see my Oblate
Spheroid Waveguides article.
Bipoles excel in high SPL capability -- not surprising, given their double-driver layout. Dipoles, on the other hand, tend to require a lot of driver and amp output capability, if you want to maximize the directivity controls and have the most consistent balance of axial and room responses. Bipoles, as with constant directivity (minimal baffle) dipole types, can present a well-balanced mix of room response and on-axis SPL. Both types excite the room to the rear of them similarly (one in and one out of phase), and as such have significantly more room involvement than an ideal constant directivity monopole speaker, but less than an omnidirectional loudspeaker. The balanced reflections of a well-executed dipole or bipole loudspeaker are considered desirable for some. A room that has a significant amount of reflected energy will often be perceived as more "spacious", and have more of a "they are here" characteristic. A monopole with controlled directivity involves the room less and tends to be more of a "You are there" characteristic.
And Some More Links
I've repeatedly mentioned AudioKinesis over the past few months, and there's a reason for that- I like Duke LeJeune's designs, finding them quite innovative. His site is not tech-oriented, but there are some interesting systems to peruse AudioKinesis.
A man who likewise should need no introduction to my
readership is Siegfried Linkwitz. A pioneer in audio and a man who uses both
omnis and dipoles. His site is a heck of a resource with many good tech-oriented
articles. Linkwitz Lab.
Another frequently dropped name is Earl
Geddes a man who has really driven some of the popularity of modern
constant directivity loudspeakers, with constant directivity monopoles like his
Summa. Some of the pioneers in this regard were Don Keele and David Smith,
working on the legendary JBL
In case nobody noticed, I'm a little bit of a JBL fanboy -- more for their efforts in making top-shelf transducers, a willingness to experiment, and a long and storied history. There are many great speakers bearing JBL's badge, not least amongst them are the modern Array, K2, Everest series loudspeakers. Unfortunately, under Harman International, they have closed much of their American operations, scrapped their production standards, and gone for the "quick buck". Stock price short-sighted capitalism at its worst. Don't mistake my love of their history and classic products with a respect for the current operations. There are still talented people and designers there, but it's been strip mined, and standards are not anywhere near what they once were.
And on that negative note, I leave you, hopefully with a better understanding of the way dipoles and bipoles work, and with an earworm planted, Wrath of Khan style. Think carefully about directivity in your speaker designs and it is one of the most important aspects of loudspeakers. Sufficient output capability with reasonably low distortion, flat on-axis response, directivity. If you're using EQ, directivity takes the #2 spot, and sufficient output capability is achievable with brute force.