Dipoles are strange beasts, and they must be given some considerations that are not evident when designing many box loudspeakers. For example, as John Kreskovsky has so aptly pointed out in his Music and Design website, he indicates that in dipoles, there is no correct compensation for vertical offset of drivers if this has not been addressed in the physical construction of the loudspeaker itself. This is due to the fact that any electronic delay introduced to correct for driver offset with regard to the front output will result in a larger magnitude of error in regard to its rear radiation. He explains this here: http://www.musicanddesign.com/Dipole-offset.html. It is a really good article, and this is a design consideration that might not be evident at all to those utilizing dipoles in their loudspeaker construction activities.
In addition to this anomaly, which is not much of a consideration in most box loudspeaker designs, there is an additional consideration, the different frequency response of drivers from the front versus the rear. I’ve seen many make the simple blanket statement that the rear response is not as important as the front. While this may be somewhat true, I haven’t seen any real studies to show the extent to which this is true. Certainly it should be a design consideration in dipole loudspeaker construction, regardless of its seeming inconsequential level of importance. For example, consider the following on-axis measurements of the absolute SPL of a tweeter I fabricated, 2.83 V/1M, front versus rear.
The black line represents the front response, and the red line represents the rear response. There is some additional steel added to the back portion of the stator between the magnets. Magnets love additional steel, as it creates an easy return path for the magnetic field as opposed to air, and this is a good way to boost sensitivity. The downside is that this additional steel makes sound waves bounce around. Physics makes its own rules, I’m afraid. We’re just here to give our own little spin on them.
Now, at first glance it would seem to be easier to just use the front response since it is much more linear and just forget about the rear response; add a little shelving to bring the bottom end up, and you’d have a nice flat response. You would end up with something like the following after equalization. The response of both front (black) and rear (red) have been normalized to 5 kHz on-axis to the front response, but with a 20 dB offset to the rear response for comparison purposes.
I would first study the rear on and off-axis response after equalization and compare it to the front response, then make a determination as to whether it was comparatively good enough to turn the tweeter around and use it in the reverse fashion. I mean, look how bright it might sound otherwise, with that extra amount of top octave sound energy from the rear reflecting off the rear wall. At the least it would produce a little extra sparkle and air to recordings that was not really there, and with a lot of today’s rather “hot” recordings, this certainly wouldn’t be necessary. How would it sound reversed? Let’s take a look at the rear response equalized for flat response instead of the front:
The response of both the front and rear now have been normalized to 11.5 kHz on-axis to the rear response since that rear response hump would probably be filtered out, and this normalization frequency represents the highest point of the broad peak. The rear (red) response is pretty good even 60 degrees off-axis up to about 14 kHz, and what was formerly the front (black) response is now depressed in most of the uppermost octave, which is probably much better than having to treat the rear wall with absorptive material to eliminate the effects of the rear response hump, if it were to be used as originally intended.
But even better, since I’m fabricating my own driver, I could make both front and rear physically symmetrical since the off-axis response isn’t bad, thereby ensuring that any equalization applied to the front would be replicated in its rear response. The response curve for both front and rear would look like the red lines in the last frequency response plot. In reality, that is what I’ve attempted to do, to keep the driver construction symmetrical front to rear. I really think it imparts a sense of startling realism to the sound, in my opinion.
Another case might be a midrange driver suffering from some of the same symptoms, maybe with a resonance in its rear passband, and here the ear is also sensitive to any response aberrations. And by equalizing the front for flat response, the same equalization applied to the rear response has now magnified this resonance. Maybe that’s the source of the upper bass “honk” you could never identify.
So what does this all teach us? Good on and off-axis response is best accomplished with magnets which are the smallest possible size and spaced far enough apart as to reflect as few sound waves as possible. What does that same loudspeaker sensitivity depend upon? Just the opposite–large and/or powerful magnets spaced as closely together as possible, or multiple magnets, or both, with much added steel that will saturate with magnetic flux easily. Stupid physics! It just doesn’t cooperate when you need it most.