Loudspeaker designer John Dunlavy: By the Numbers... Page 5
Dunlavy: Yes, it does. In an antenna, generally you want the beam tailored so you get the energy directed in a certain direction to achieve maximum efficiency. Similarly, if you want a speaker to have a fairly flat room response, you can only achieve that by tailoring the design of the speaker such that the separation of the various drivers, as expressed in wavelengths across the audio spectrum, maintains a more or less constant vertical beam-width as well a fairly constant horizontal beam-width. That comes about by choosing drivers of different diameters. Say you have a 15" woofer, and you measured the beamwidth at 1kHz, then you compared that with, say, a 4" midrange at the same frequency. Good heavens—no comparison. A 4" midrange will have a very broad beam; the 15" woofer will have a very narrow beamwidth—you can find it by dividing 57 by the dimension expressed in wavelengths. That's the beamwidth.
You want symmetrical coupling into the room, which then implies symmetrical radiation patterns from the speaker in both horizontal and vertical planes. To achieve that, you have to have a symmetrical driver array. You place the tweeter at the average height of a seated listener's ears, which we take to be between 37" and 40", depending on the person in the chair. Then the rest of the drivers—a pair of midranges and a pair of woofers, or whatever—are symmetrically arranged above and below that tweeter.
Atkinson: In one sense the SC-VI is the end of the line. The next Dunlavy Audio Labs flagship loudspeaker, the Magnus, is a DSP-corrected design. What are you trying to achieve with the Magnus that you haven't already achieved with the more conventional speakers?
Dunlavy: You can use digital components to achieve extremely accurate time-alignment of the drivers. In the Magnus, we are able to time-align the drivers to an accuracy better than 1.9µs, which is an exceedingly small distance, less than a sixteenth of an inch. And we can maintain that over a very, very wide frequency range.
Atkinson: Whereas you achieve the time-alignment physically with the SC-VI and the smaller conventional Dunlavy designs by recessing the shallower drive-units, the Magnus has a flat baffle.
Dunlavy: That's right. You can also, in the digital domain, separate the amplitude components from the phase components. You can only do that digitally. With the Magnus, we can individually EQ each of the peaks and valleys in the response. We can achieve a frequency response approaching ±0.25dB. We always start off with a speaker that is at least better than ±2dB, so the digital part of the system doesn't have to do a lot of work. The more work it has to do, the more chance of error creeping in. So we try to start off with a loudspeaker that's probably already better than anything else out there, then we let the digital just wipe up the remaining mess. We wind up with a speaker system that's near-perfect.
The Magnus incorporates a true class-A, true push-pull FET amplifier on each driver. We have a 200W amplifier feeding the tweeter, a second 200W amplifier feeding the midranges, and the third 200W amplifier feeding the woofers. Because the drive-units are directly connected to the amplifiers, we get the additional damping factor due to the very low internal impedance of the FET power amp.
But we think ultimately by being able to work on optimizing the interaction of all components—the drive-units, the amplifiers, the D/A converters—we can assemble a system rather than just a speaker that may be near-perfect.
Atkinson: What's the downside of going the digital route? Presumably you have to have very good DSP and DAC chips to avoid introducing quantizing noise and other errors.
Dunlavy: And, in fact, that's the reason we don't yet have the Magnus to market. The current version employs 16-bit digital chips. While our feeling is that 16 bits is terribly close to being acceptable, being devoted to perfection we want to go to market with at least 20-, and preferably 24-bit chips. I think 24 bits probably means we will have about as close to distortionless performance as one would ever be able to wish for. We also think that the system should be a 24-bit system at a very high sampling rate so that if someone buys a Magnus two or three years from now, it's not going to be outmoded by better D/A converters and whatever.
Atkinson: I assume the Magnus will be expensive.
Dunlavy: We have no idea at the moment what the selling price for the Magnus will be. It obviously won't be affordable! The cabinet alone costs several thousand dollars—at our cost! It's like an SC-VI except that, with its rounded side edges, it's far more complex. It's an absolute nightmare to build. Total nightmare. Enormous internal bracing. We designed it largely by computers, calculating where to put braces to null out internal modes. There are three—sorry, four—I'm sorry, five separate internal enclosures housing the drivers. So, it's just an incredibly complex nightmare of a thing to put together.