Design guru Andy Munro talks to PAUL WHITE about the development of Dynaudio Acoustics' monitors and his own approach to studio design.

ACTIVE VS PASSIVE

Could you explain the advantages of active systems over traditional passive designs? I've always understood that a good active system provides you with more control over the crossover filter characteristics, and you get better coupling between the amplifiers and the drivers.

"You get all of that, but a well-designed passive crossover can actually better complement the impedance characteristics of a driver than an active system can. With an active system, the amplifier drives directly into the loudspeaker, but a driver in itself is a fairly complex load -- the impedance rises rapidly with frequency; it has dynamic and mechanical impedance; it has all kinds of things going on that can't be compensated for with a voltage amplifier. You can do the job with a constant current amplifier, where the loudspeaker acts as a load -- a transmission-line kind of principle -- but most amplifiers don't see an ideal load from a drive unit, so it's not actually true to say that active systems load loudspeakers better. A passive system can actually take into account the characteristics of the driver, so you end up with a much flatter impedance curve.

 

 

"What partly bears that out is that the very best loudspeakers in the esoteric hi-fi world are nearly all passive. The top three hi-fi speakers in the States are all passive -- and just happen to use Dynaudio drivers. You can argue it either way, and we play devil's advocate to some extent in that we now look at it from both points of view and offer the choice of an active system or a passive system. I'm not saying our active systems are better than our passive systems -- I'm just saying that they're the best active systems we can make."

INTEGRATED ACTIVE MONITORS

What improvements have you been able to make in the smaller active systems, where the crossover and power amplifiers are in the same box as the drivers?

"A lot of it is trying to develop a better output stage to deliver the kind of sound quality we need. We did a lot of looking around to find the best-sounding MOSFET devices, and then designed backwards from that. Then we designed a power supply that could provide the kind of current and voltage characteristics we wanted, and, finally, we tried to put it all together with some active crossover and equaliser circuits that made the best use of the amplifier power. That allowed us to be a little more creative with the box tuning and the overall EQ of the drivers. We've gone for the best acoustic loading of the driver in the smallest possible box, with the best equalisation characteristic to go with it. This has allowed us to make the box slightly smaller than we would have done had it been a passive system, and this in turn results in a slight kicking up of the Q_tc [total resonance] of the system, which means you get a slight bump in the response. We've equalised that to be flat, which in effect gives us more headroom in that key kick-drum, low-snare area of the spectrum where a loudspeaker works hardest. It gives around 6dB more headroom in that region than the other active systems we've tried.

"We've also been able to extend the low-frequency response a bit more. Instead of using a fourth-order filter to chop the bottom end off viciously, we've allowed it to drift down a little bit lower, then used an electronic filter which is variable. This lets you work with lower power but extended bass, or you can tighten up the bass and use more power. The EQ circuit is actually quite complex."

TIME AND PHASE

"The phase correlation between the porting, the electronics and the drivers is as good as we can get. It's not necessarily absolutely linear all the way down to the bottom, but few loudspeakers are. Most phase responses are only measured from 100Hz upwards. Some speakers I've come across are very woolly or woofy at the low end -- they look as if they've got a good response, but if you hit them with any dynamics in the real bass, you don't get a click, you get a boom. I don't like knocking other people's products, and there are lots of things about other people's products I like, but if you take the B&W 801, which is a lovely classical listening device, put a kick drum through that and you can go and make a cup of tea while it's still coming out. That's why you don't see many 801s in rock-and-roll studios -- what makes it a great classical speaker might make it entirely unsuitable for anything else.

 

"The BBC used to do it with roofing felt."

 

"I'm not going to stand up and say we're right and everybody else is wrong, but we do know what we're talking about. I know what's good and bad about both our systems and our competitors'. If I had to say what was bad about ours, I'd say they weren't particularly efficient. There's a reason for that: it's because we're using certain types of technologies that aren't electro-mechanically efficient. We use well-damped, fast, lightweight components, and we still feel we have the best compromise between fidelity and efficiency. All Dynaudio drivers are matched to within half a decibel, so there's no need to match individual pairs of components.

"There are systems that are louder than ours, and there are systems that are better than ours in terms of absolute fidelity, but you can't have both. For example, look at horn loudspeakers. They're great for getting level, but in absolute fidelity terms they're diabolical. A horn is a band-pass filter, and a pretty poor band-pass filter at that, so you have to equalise them heavily to make them behave. You can't get a horn that has a constant power bandwidth, so you have to make all kinds of compromises to the horn, the driver and the equalisation. You may end up with something that sounds OK on axis with certain material, but if you put opera or piano on it, or even a flute, you can hear it. Listening to flute on horn loudspeakers is particularly revealing. A flute has very few harmonics and those that are there are very even. Over a horn, you hear all those third and fifth harmonics -- and they're substantial."

FILTERING

When it comes to studio control-room design, how valid is it to use EQ to try to counteract problems that are due to the acoustic properties of the room?

"Equalising a speaker to correct its response is OK, but a good speaker shouldn't need very much EQ anyway. However, the interface between the room and the loudspeaker can create the need for equalisation. For example, if you take a speaker designed for free-field operation and you put it in a wall, in theory you get a 6dB low-end increase on axis at, say, two metres away, so you may want to correct that. That can be a good thing, because once you've put in that EQ, you have more headroom.

"You can't, however, using normal minimum-phase equaliser filters, correct for most of the effects attributable to the room, because the majority of those are caused by reflections. If, for example, you get a dip in your monitor response created by a reflection, then you put more energy into the speaker at that frequency to try to compensate -- but you're also putting more energy into the reflection. Trying to equalise out a big notch in the response is probably going to make it worse, and that's where people go wrong. If you were to use a third-octave graphic equaliser, you might be able to get what looks like a flat response, but all you've really done is to use the filters either side of the hole to pull up the energy, and what they're doing is masking what's probably now an even deeper hole. The dip is probably deeper after equalisation than before, but the only way you can really see that is with an FFT analyser or some form of swept-tracking analyser. That way you can play around with your equalisation and see which peaks and dips respond correctly."

Is it still fair to say that your ears don't respond in the same way as measuring microphones, and if you EQ the speaker too much, the direct sound will be perceived as wrong?

"It depends on the situation, and that's where design experience counts for a lot. I think what you're saying is that the direct-to-reverberant ratio of what you're hearing is weighted in favour of the direct sound. This is true -- what you hear first, you pay more attention to -- but that effect diminishes as the frequency gets lower. At very low frequencies, you hear very little direct energy from the loudspeakers -- usually less then one per cent. Even with a small nearfield monitor, the surface area of the radiating bass wave is 13 square metres just one metre from the cabinet. The area of your ear, on the other hand, is about five hundredths of a square metre, so the amount of energy you hear directly from a loudspeaker in a normal room at low frequencies is a fraction of one per cent. All the rest comes from the room, unless the room is completely dead, in which case you'd have a system that's hopelessly inefficient.

"My particular argument against completely dead control rooms is not because I don't think they're accurate, but because I have an inbuilt aversion to wasting all that energy! If the room is acoustically correct, you can make use of that energy, and even in a fairly dry control room with a reverb time of about a quarter of a second, you can still get somewhere between 6dB and 8dB more out of a monitor system without actually compromising the deadness and the accuracy of the room at all. We very rarely find any need to EQ other than at the low end -- most of our systems are inherently flat in a reasonably correct room.

"We try to get this balance between the direct response of the speaker and the room response, which means the off-axis response of the speaker has to be quite good. And that's where I have objections to some of these so-called time-aligned monitors where they just shove the tweeter back a few millimetres. That's only time-aligned on axis, but what's needed is proper phase alignment of the drivers such that the phase response is correct both on and off axis. Again, this is where a passive system can be an advantage, because you can play around with the phase alignment to make it work.

"All this was worked out by Richard Heiser, who also invented the TEF analyser, and he published AES papers in the late '60s and early '70s where he described how minimum-phase filters worked. He described how any system with a flat frequency response would have flat phase response, if you use minimum-phase filters. But you can make a system look flat without using minimum-phase filters, and it will be all over the place. That's where the time-alignment concept falls apart. I've come across a lot of systems where the frequency response looks fine, but the impulse response doesn't look at all right. That's why most loudspeaker manufacturers don't publish phase data, or distortion data, or group-delay data -- it's misleading and, even in the best systems, it never looks that great."

THE ROOM INTERFACE

"If you have too much high-frequency reverb in a room, it's not going to affect what you really measure in terms of frequency response on a particular loudspeaker system -- it'll just sound harsh. We've found that we tend to let the high-frequency response tail off a bit on most systems, and it works. But the systems that have the widest dispersions at high frequencies can stand the flattest reverb characteristics. In other words, if you have a speaker that's flat to 20kHz, with a perfectly dispersive polar response at 20kHz, then you can have a room that is actually quite reverberant at 20kHz, and it will all work. If the room starts to dry up in the high frequencies, then it's better if the speaker-directivity tightens up as well. Intuitively, this sounds like a contradiction, but it's not.

"The perfect example of that is the cinema. If you've got a great big JBL system and you're measuring 20 metres away, then you have to put in the X curve or the Academy curve, which is a roll-off above 2kHz. The reason for doing this is not, as many people think, to tone down the horn, but to correct for the acoustics of the room. The room is much deader at high frequencies than it is at low frequencies, so it's really compensating for that. You actually get a build-up at low frequencies -- but the only energy you get from a horn is what comes out of that horn so, allowing for the inverse square law, you get the same amount of overall energy at 20 metres as you do at one metre. However, at low frequencies, most of the energy goes back into the room, and a very high percentage of that comes back to your listening position, which increases the bass level. To maintain the correct balance at that point, you have to allow the bass to be much higher in level than the high frequencies, to maintain the original balance. I had a lot of problems understanding that, and it wasn't until I worked it out mathematically that I understood it properly.

 

 

"If you listen to a pair of flat nearfields at one metre, and compare that with the more distant main monitors having the appropriate roll-off, they should sound the same, even though the nearfields may be flat to 20kHz, and the more distant monitors may be down 15dB at 20kHz. It's all to do with the direct-to-reverberant ratio. There's a lot of apparent contradiction in this, and if you look in some of the textbooks, they'll tell you that you should increase the reverb time at high frequencies if the speaker has a narrow HF dispersion."

ROOM DESIGN

What is the latest philosophy on control room design? This seems to change from year to year...

"I like a neutral room with no obvious characteristics, but it doesn't have to be dead. Probably the biggest challenge we've had recently is doing the remix room up at Decca when they bought an AMS Logic 2 and went completely digital. They discovered that their optimum listening environment has a slightly longer reverb time than we would normally have done, about 0.35 seconds, but extending right out into the higher frequencies. Now they use the B&W monitor, which has a tiny tweeter with a very wide dispersion, so this is where we came to this conclusion that if you have a very extended high-frequency dispersion, then the high-frequency reverb should be there to match it.

"Most of the best-sounding rooms around are pretty average in terms of what I call space technology. They are well designed and have properly balanced reverb times, but they don't use anything particularly extravagant."

Do you advocate the use of diffusers in your designs?

"Oh yes, we use them a lot, but they're just part of the palette of available treatments. If you want to get rid of early-order reflections, and you don't want to the room to be too dead, then diffusion is the best way to do it. The theory is that reflections are scattered over a hemisphere but, in practice, the diffuser just trashes the reflection! One thing that people tend to forget, though, is that diffusers also have quite a high absorption co-efficient, so you have to take that into account when planning your overall room absorption. A lot of concert halls that were done in the '80s used diffusers, and in many cases they had to take the diffusers out, because the reverb time ended up being too short. To some extent, that can be true in a large control room."

The traps you're using around this room, based on shallow boxes covered with a membrane of Revac deadsheet, seem to be a relatively recent innovation.

"Yes, but the BBC used to do it with roofing felt! Traditional bass traps are a bit crude, and a quarter-wavelength bass trap is just that -- which means it has to be very deep to get down to, say, 20Hz. The trap we use now is based on the traditional panel trap, but the damping factor is different. There is another term in the formula, but basically it's still a mass on a spring. You have a certain mass of material, and a certain air stiffness in the box behind, and it creates a natural resonant frequency. The advantage of this design is that at that natural frequency, the movement is maximum, and that results in a loss of energy. The stiffer the panel, the narrower the Q; but with this very dead material, we can create a low-Q trap that works over quite a wide frequency range with a Sabine co-efficient of about 0.7. We use three types: the deepest, which is around eight inches deep, has a maximum absorption at about 40Hz. The others work at about 80Hz and 100Hz, but we tend to stagger them to try to give overall absorption over a wider range. We can also make them with a higher Q by bonding the Revac onto very thin plywood. That works quite well for tuning out room modes.

 

"I don't think the live-end, dead-end thing was ever valid."

 

"There was a studio in Camden Town built in a dead square room, and it had the most wicked 40Hz modes of any room I've ever seen -- if you looked on an analyser, the 40Hz band was around 12dB higher than the rest of the spectrum. What we did in there was design these boxes to resonate at 40Hz with the narrowest Q we could get, and we tuned that mode out altogether. The other way to do that is to use an electronic notch filter in the monitoring chain to put in a very deep notch over, say, a 3Hz or 4Hz band. You can notch out a low-frequency mode and it makes practically no difference to what you actually hear. Theoretically, that's an awful thing to do, but it's just knowing what to do in the circumstances."

In terms of whereabouts the absorption goes in a room, do you now advocate distributing it evenly around the room, rather than going for some type of live-end, dead-end concept?

"I don't think the live-end, dead-end thing was ever valid. I think it was just a misconception to start with because, at very low frequencies, there's no such thing as live-end, dead-end. Distribution at low frequencies is modal, and therefore by definition colours the whole room. You get to a transition point, known as the critical frequency (the formula was discovered by Manfred Shroeder), which is a function of the room volume and the reverb time. For a typical small control room with a reverb time of a quarter of a second, the critical frequency is around 150Hz. Below 150Hz, the room is entirely controlled by its volume and its modes, and above that, it's more or less the specular reflections and the geometry of the room that influence its behaviour.

"With our bigger systems, we often use a sub-bass system to control the bass end, then cross over into a normal system, where we try to match the crossover point of that system to the critical frequency of the room. In a bigger control room, that would be around 100Hz, and I think you can get terrific results by using sub-bass up to 100Hz, and then having a satellite system to do your main monitoring."

In the small studio, what's the best way to treat the back wall? Should you use absorbers to try to make is disappear, or is it best to use scattering?

"A bit of both, I think. It's useful energy, so if you can diffuse it and maintain a reasonable reverberation time overall, then that's a good thing. What's bad is to create a live-end, dead-end feel, so that when you're at the back of the room, you do feel that you're in a different acoustic environment to the mix position. There are some studios where the engineer is hearing a more or less dead, direct sound, and the producer is hearing the reverberant characteristics of the back of the room.

"In the very small studio, it can still be a valid approach to put a Revac trap across most of the back wall and then cover this with foam to provide broad-band absorption. But it's not too difficult to get the energy elsewhere rather than absorbing all of it. Even if you've got a relatively shallow room, providing it's reasonably wide, you can deflect the energy onto the sides, then diffuse it, then bring it back. We tend to do more of that rather trying to make it completely dead."

What would be the practical implementation of that? Would you build an absorbent wall and then put diffusers over the top?

"Basically, a low-frequency absorbing section with a diffusing or reflecting surface on top of that. Or we might actually use a veneered ply as the basic absorber, but done in such as a way as to deflect the high frequencies."

How has the science of acoustic treatment improved over the past couple of decades?

"Measurement is the key to everything. There's only one real aspect of acoustics that I think has improved over the last few years, and that's the measurement side of it. Trial-and-error room-tuning worked pretty well, and most people who knew what they were doing could get a room to sound OK in the end, but nowadays there's so much more that you can measure, analyse, and deal with on a computer. I don't think there's much else new in acoustics that we didn't have 30 years ago."

"We've used a variety of amplifiers over the years, including a version of Malcolm Hill's design that he built especially for us, and we've used the Chord amplifier. The Chord has the dynamics we need, but it also has the kind of power supply that can deliver high levels of current when required. Even with the passive systems [such as the BM5 and BM10, reviewed in SOS June '96], it's surprising how dynamic they can sound when they're driven from the right amplifier, whereas nearly all the smaller active systems I've heard just run out of steam. They use various devices, such as compression, limiting, or signal processing, but none of them can match the dynamics that you get from using a big amplifier. It's all down to money.

"I think that active systems fall into two camps. You have those designed for workstations, multimedia, home studios and so on, where they don't have a lot in the way of dynamics or headroom, then you have what I call the really well-designed professional systems that have been around for a number of years, from the likes of ATC, and Genelec, Meyer HD1s. If we were going to make an active system, we had to do something different."

Have you used DSP controllers to provide more elaborate control over loudspeaker systems?

"We've used DSP technology to improve the driver/room interface and, as DSP gets cheaper, I see it being used in more products. The Sigtech uses 24 DSPs and costs around £5000, and you have to have high-quality converters, but then that isn't a huge amount of money if you have a room that has problems, and DSP allows you to sort it out quickly and effectively. We've even, on occasions, included the Sigtech system as part of a monitoring package, so that when we come to equalise the system we can do what we want. As well as the FIR filters in the system, you can actually put in your own house curves to get the system just the way you want it. There's a lot of mileage in that."

"We use a couple of proprietary modelling systems -- one's called CAT acoustics, which basically is a way of developing algorithms that can drive a DSP engine to create an acoustic impression of a system. It's crude, but it lets you listen to the room before you've built it. The degree to which you can model something is almost infinite, and you have to do it to an almost infinite degree to get a realistic model. The basic problem is time -- time to input data in the first place, and time to process it afterwards. Most acoustic systems at the moment work on co-ordinates, so you have to put in a wire-frame drawing of the building, then you can define surfaces. It's very good for designing sound systems for large venues, but less accurate for modelling the behaviour of small rooms."

 

Home | Search | News | Current Issue | Digital Editions | Articles | Forum | Subscribe | Shop | Readers Ads

Advertise | Information | Links | Privacy Policy | Support