Reference

EUREKA Archimedes
Somewhere in Denmark a research project is in progress to determine the subjective effect that the listening room and loudspeaker directivity have on reproduced sound. More specifically, under a European R&D funding scheme called Eureka, Bang & Olufsen of Denmark, KEF Electronics Limited of the UK, and The Acoustics Laboratory of the Technical University of Denmark have embarked on a project called Archimedes. They hope to find, first of all, if it is even possible, given current technology and funding constraints, to arrive at useful specifics and generalizations regarding the room/loudspeaker/listener interface. The ultimate hope is that a "smart" loudspeaker may be possible, one that will compensate automatically for its surroundings. Even a less ambitious outcome could be extremely useful. Manufacturers today cannot even agree on the appropriate directivity pattern for a loudspeaker, and studies which purport to tell them which is "best" still rely on a lot of educated guesswork.

There is no guarantee that Archimedes will come up with useful data, though the strong probability exists that at least some worthwhile results will emerge. The project has been set up in the anechoic chamber of the Technical University of Denmark, which, at 26' high, 33' wide, and 39' long is one of the largest such facilities in Europe. It is effectively anechoic to below 50Hz. The initial portion of the study is designed to study timbral effects, and the simulation is therefore at present limited to a single loudspeaker reproducing an anechoically recorded monophonic signal.

In order to understand what is involved in this simulation, visualize a single loudspeaker suspended, oh, a mile above ground and a (preferably non-acrophobic) listener similarly suspended. Play back a signal through the loudspeaker and---absent the stray passing 747 or hang-glider---the listener will hear, effectively, a reflectionless or anechoic playback of the chosen program material. Given appropriate control of the signal's frequency balance, a blindfolded listener will be unable to accurately determine the distance of the source; a very nearby source played softly will seem to be at the same distance as a distant source played loudly. The surface reflections which provide us with critical cues to distance will be totally absent.

Now add a single wall a few feet to the listener's left---levitated in the same fashion. The listener will now hear not only the direct sound, but the sound from the loudspeaker as reflected off this single wall. But the latter will not be the same as the direct signal; it will, of course, be delayed in time, but in addition its frequency balance will depend not only on the frequency-related dispersion of the loudspeaker (which will determine the frequency balance of the sound at the wall just before reflection), but also on the reflection characteristics of the wall---also frequency-dependent.

Now add a second wall and you can add not only the reflections from the two walls, but any reflections which traverse both walls in turn before reaching the listener. You can see that this is getting awfully complicated awfully fast. Bring our intrepid astronaut-listener-in-training back down to earth and surround him or her with the usual six walls of a typical room, and you can begin to grasp the complexity of the problem that the Archimedes researchers have chosen to grapple with.

In order to attempt to simulate a real-world listening room, the listener is positioned inside the anechoic chamber at the center of a low-reflection, spherical gridwork three meters in radius. A large number of two-way loudspeakers are positioned on this sphere according to the room being simulated. The listening chair is motor-controlled so that the listener's head can be precisely located. Six of the loudspeakers are dedicated to simulating the late-arriving reflections which are generated by means of a multi-channel reverberator. The early reflections, those occurring within the first 30ms, are considered to be the most significant. After making several simplifications based on experimentally established psychoacoustic properties (one of the most noteworthy being localization blur, which enables one loudspeaker to represent multiple images coming from the same general direction), digital signal processing is used to generate the signals to be fed to the loudspeakers reproducing the "early arrival" reflections. The required signal-processor array, its associated interfaces, signal-routing hardware, and controlling software are collectively referred to as the "DSP Engine," without which this project would not be possible.

The loudspeakers used for the simulation incorporate coaxial two-way drivers---similar to the drivers used by KEF in their commercial UniQ line. The selection of such coaxial drivers was not incidental; spatially separated woofers and tweeters add an additional, unnecessary complication to the simulation process. Each driver is mounted in a spherical cabinet, the best shape for minimizing diffractive problems. The cabinets were originally designed as net floats used by Danish fishermen!

I was privileged to visit the Archimedes project last year along with a number of other journalists. At the time, the "DSP Engine" had not yet been put into the system; it had only just arrived at the university. For the preliminary stages of the project a less complex series of time delays and equalizers were used just to get the system up and running. That was the configuration we heard. Each of the journalists took a turn seated in the "hot seat." When I took my turn, what I initially heard was an anechoic cello recording played back near the tip of my nose. Or so it seemed, though I knew that the loudspeaker was actually several meters away. (An acoustically transparent screen surrounded the listening chair, hiding the exact loudspeaker positions and the rest of the chamber environment from the listener.) When the "room simulation" was turned on, the cello moved back about 10', surrounded by a relatively natural room ambience and sense of space. The overall fidelity of the reproduction was only fair---at that stage of the study, fidelity was not the point. But it did indicate that, even without the sophisticated digital signal-processing equipment which by now is in place, the combination of chamber, critically placed loudspeakers, and processing could conjure up a believable sonic environment.

Archimedes will not be finished this month or next year. But at last an organized attempt is being made to apply current technology to the arcane subject of loudspeakers in listening rooms.---TJN