Robin Marshall: A Modicum of Genius Page 6

The woofer is very directional at the crossover frequency, while the tweeter is almost omnidirectional. You've got that horrible, sudden transition. So if you can spread the transition between the two drive-units over a larger part of the bandwidth by using gentle-slope crossover filters, you can fudge it, I suppose is the best term.

Atkinson: You would agree then with Floyd Toole's findings that, off-axis, you need a controlled dispersion across the band, with no sharp discontinuities at any frequency?

Marshall: Yes, absolutely. However, what you also get with gentle-slope crossovers is that when you start to measure the box, you can find all sorts of problems, particularly when you move your microphone in the vertical axis. You get nulls which can look fairly horrific. And if you choose to, you can get measurements of the box which can prove beyond a shadow of a doubt that it is a piece of junk.

Atkinson: But these nulls are due to cancelations between the radiation from the two drivers having different path lengths specific to that point in space, not to an actual lack of energy at that frequency.

Marshall: It's a totally false problem. But you tell that to the measurement-oriented people who say, "Oh, look at this cancelation."

Atkinson: First-order, 6dB/octave-slope crossovers have always been popular in America. First, they are time-coherent. Second, they avoid the necessity of having to wire the tweeter 180 degrees out of phase to get rid of the cancelation notch at crossover with 12dB/octave slopes.

Marshall: Third, they are bloody cheap!

Atkinson: What would be your reasons for using simple, low-order crossovers?

Marshall: The ones we've just talked about, the polar response changes, also the transient response of a first-order network. If you've got to have a roll-out, then choose the gentlest one so you don't upset the time-domain response too much. A first-order network is the only network where you haven't got any time-domain problems. There is no other, although people are always developing trick networks, aren't they? You know, "We really licked the problem here, this is really it." And some of the stuff that people like Stanley Lipshitz have come up with, they claim to have overcome these problems, but they never sound any good to me.

That's always the final arbiter to me. I don't give a pat what the theory is; if I listen to it and it sounds awful, then it's not working, whatever the theory tells me. Like I said at the beginning, since my education was in mathematics I'm not ignorant of the theory, but a lot of people accuse people like me, people who trust their ears sometimes more than the theory, saying "Oh well, that's only because he's such an ignorant bugger, you know."

Atkinson: You've gone beyond just using low-order crossover slopes in the '14. There actually is no low-pass filter in the bass-unit path. Doesn't that put severe constraints on the drive-unit?

Marshall: Yes. It means that I have to design the bass driver to perform its own roll-out slope. Again, there's nothing new, no magic about doing that. It's purely an engineering problem. By adjusting the various masses of the components, you can put the roll-out slope wherever you want it. Of course, you can't make the woofer roll out at 6dB/octave. By its very nature it wants to roll out at 12dB/octave. All you can do is make the roll-out slope very low-Q at the very early part of its roll-out, so it looks like a 6dB/octave slope to begin with.

Atkinson: It achieves its final 12dB/octave dive when you're an octave or so further up?

Marshall: And then, of course, it's going into total, uncontrolled breakup beyond that, and the roll-out is God knows what. All you can do is to try to engineer all those things as far away from the passband as possible so that they don't matter. In reality, of course, they do matter. I mean, any resonance is audible; even a resonance 30dB down will still be audible. It hasn't gone away because you can't see it on the measurement.

Atkinson: A final question: you obviously think it important that anyone involved in designing hi-fi components should have some contact with the real thing. You yourself are a musician. How do you go about organizing your listening so that you know that you're picking up things which are real?

Marshall: I don't make any attempt to really. The only way I discipline myself is to make sure that I listen to an awful lot of real music, both serious music and fun music, like jazz. I listen to all kinds of music. And I do that in a lot of live environments. I don't do it as a chore, I do it because I enjoy doing it! And when I listen to equipment, I listen to it in the same way that I'd listen to a live concert. I don't go to a live concert and start analyzing it, saying, "Hmm, the highs are a bit gritty." I sit down and take in the overall sound and say, "Am I enjoying listening to it? Is this moving me? Is it doing things for me?"

Too many people, particularly when they go to buy hi-fi equipment, are so on edge listening, trying to pick out certain parts of the system's performance, that they never take in the whole! That's what's important. That's what music is about.

There's a question I think your readers should always ask themselves when they sit down to listen: "Am I enjoying this?"