Jim Thiel: A Coherent Source Page 3

Atkinson: One of your goals appeared to be reduced woofer distortion. You reinvented the idea of using a copper shorted turn on the magnet pole-piece.

Thiel: What that does is stabilize the strength of the magnetic field in the gap. The ferrite permanent magnets that almost everyone uses are not perfectly stable under load. And when you run many amps of current through the driver's voice-coil, the magnetic field it generates pushes against the permanent field of the ferrite magnet and actually demagnetizes it to some degree. So the magnetic reference point that the voice-coil is interacting with is constantly changing. This results in power compression and also in distortion.

By incorporating a heavy ring of copper around the central pole-piece of the driver, you can reduce dramatically the changes in the magnetic field strength that are induced by the current from the power amplifier. You get a more dynamic, less compressed result, with lower distortion. So the bass reproduction is more tonal and has more impact.

Atkinson: One of the things you've done to reduce distortion has been to go from a conventional voice-coil topology, where you have a long coil that overlaps a short magnetic gap, to the opposite, where you have a very short coil operating in a long magnetic gap. What are the advantages of that?

Thiel: The entire coil is in the very uniform magnetic field within the gap at all times. And since the intensity of the magnetic field from the front to the back of the gap changes very, very little, and the entire coil is in that gap even during long excursions, the distortion is very, very low. You get a truly dramatic reduction in distortion. The disadvantage is that since you have a much longer magnetic gap, you need a much larger magnet to power that longer gap, and therefore, everything being equal, it will be a more expensive driver. But the advantages of the distortion reduction are really substantial.

Atkinson: I understand you've gotten heavily into modeling the magnetic circuits of your drive-units.

Thiel: Yes. By using finite-element analysis on a computer, we're able to actually simulate what the magnetic field distribution is in a hypothetical magnet system. So instead of actually having to construct an actual magnet system and do tedious measurements on it, we can run hundreds of experimental simulations on the computer to develop magnet-system geometries that have much more uniform and symmetrical magnetic fields and therefore have much lower distortion. Merely by changing the geometry of the center pole-piece and the magnet system, you can reduce distortion by an order of magnitude, to a tenth of what it otherwise would be.

It's really quite exciting, because these changes in the design of the driver cost virtually no money. You're tooling this center pole-piece on an automatic lathe anyway, and it makes very little difference what shape you're turning the metal into. But it makes a dramatic improvement to the behavior of the driver. The distortion that is introduced by the magnetic-system nonlinearities is no longer the determining factor in the overall driver distortion.

Usually in woofer design at higher output levels, the magnetic-system nonlinearities are by far the largest source of distortion. In our new woofer designs, this is no longer true. Distortion is introduced by minor suspension nonlinearities and by inductive effects in the voice-coil and its former, and even resonant behavior in the diaphragm becomes more significant than the distortion introduced by the magnetic system. So magnetic-system distortion is virtually eliminated.

Atkinson: But to see your theoretical ideas worked in practice, you were still ending up with a set of specifications that you sent to your drive-unit supplier for them to make and for you to then test?

Thiel: Yes. We would send them dimension drawings, tell them exactly how to make all the parts in the magnet system. Then they would build up samples for us and send them to us for evaluation.

Atkinson: How many iterative loops did you typically go through with the drive-unit supplier?

Thiel: In the early '80s it was probably only two or three. By the early '90s it was probably a dozen . . . it could take up to a year to optimize a driver design. If I wanted to try a different diaphragm profile or a different magnet-system geometry, from the time that I requested samples from them until I received the samples to test could easily be a month. So this would severely limit the number of experiments we could run to try to get a new product out. Most recently, we've started manufacturing our own drivers so that we can implement designs that are more exactly what I desire.