Features

This position is philosophically untenable: to involve the concepts of "accuracy" and "realism" must be fallacious when discussing just one component in the reproduction chain, because you then introduce another variable: the quality of program material. If you are to assess the imaging properties of a pair of loudspeakers using music program with subjective reference to the live experience, you must know how well the stereo aspect of that live experience has been preserved on your two information channels. But you only have access to that information by listening to the program on loudspeakers of unknown imaging properties. You have one too many variables—and a circular argument.

It is also untenable on more practical grounds. Let's look more closely at the problem of identifying sound-source location in the concert hall. As a regular concert-goer I am very much aware that my degree of confidence in aurally locating instrumental images varies enormously. It depends on the hall, on one's seat position, the presence of reflecting walls near the instrumentalists, the nature of the instruments themselves, and so on. Take London's Festival Hall. Sitting in the centre about halfway back, so that the orchestra subtends an angle of about 60°, one can accurately locate most instruments—apart from the horns, which have their bells pointing towards the rear wall.

As you move further back, however, two things happen. Firstly, the angle subtended by each instrument gets smaller, thus increasing the relative effect of any confusing element, and secondly, the ratio of reverberant to direct sound intensity increases. In the Festival Hall, as one moves back under the balcony there is a marked change and it becomes much harder to locate individual instruments with any confidence. The direct wavefront from the oboe, say, now only represents a very small fraction of the soundfield at your listening position, and it is not surprising that you can't be absolutely sure of the position of that oboe.

This is the situation described by Hirsch and many others. But consider the soundfield at that point more carefully. Your ears are still receiving a mass of directional information which is being interpreted unambiguously: although most of the sound impinging on your ears represents reverberation, your brain acts on those reflected wavefronts in exactly the same way that it does on direct wavefronts. This is how you know that you're sitting in the Festival Hall, near the back under the balcony, with a relatively narrow orchestra unambiguously in front of you. Your direction-finding apparatus is as accurate as ever, but by choosing to sit further back you've reduced the proportion of useable information concerning the individual instrumental positions; in effect, you reduce your brain's signal/noise ratio when you consider the orchestra alone.

But what does this relative lack of orchestral directional information have to do with loudspeakers? At a recent AES meeting on microphone techniques (see HFN/RR March 1981, p.37) Peter Baxandall put his finger on the crux of the whole business when he said that he preferred to sit so far back, because although the orchestral positioning accuracy was impaired the overall sound was better, being better integrated.

Choosing to sit that far back is purely a matter of personal taste. A loudspeaker which confuses the location and width of our infinitely narrow central double-mono image, and hence the locations of the whole lateral continuum of narrow images (exactly as if we had been removed several rows of seats further back in the Festival Hall) is a less good design in high fidelity terms because it is adding a distortion to the program. The fact that many would find the effect of that particular distortion more like the realism they prefer, regardless of what directional information was captured on the recording, is totally spurious.

The narrowness and stability of the central image at all frequencies, and the maintenance of that narrowness (within established psycho-acoustic limits) at all positions across the stereo stage as the L/R voltage ratio is altered, is the only certain indicator we can have of the absolute imaging properties of a loudspeaker. The fact that many records will sound less attractive to many people with such a loudspeaker can only be a criticism of the quality of the program.

Accuracy vs Musicality
Which brings us to the split becoming more and more apparent amongst enthusiasts and manufacturers alike. If most records sound less-and-less to your taste with every improvement to your system, then why bother with the whole business at all? Paul Benson, ex-Editor of Hi-Fi Answers, highlighted that particular point in a stimulating if controversial article last September. "The goal of a stereo system is to create, in the home, a musically satisfying performance" he wrote, and if the more pedantic critic will be upset by the notion that one's hi-fi might be creative as opposed to reproductive, it is surely true that unless possessed of masochistic tendencies one does basically want to enjoy the music on one's records.

Hence the appearance of loudspeaker systems which sacrifice the stereo aspect of performance to a greater or lesser degree, dependent on the acoustics of the listening room, in order to optimise other aspects. The Linn Isobarik, Sara, and Kan, Heybrook HB3 and ARC 101/202, are all designs which are intended to be stood against a wall to achieve their intended bass loading. However, the reflections from that wall follow the initial wavefront after so short an interval, around 2–3 ms, that the ear/brain tends to treat them as part of that wavefront, and as the level of such reflections will be dependent on frequency, the overall effect is unpredictable and results in a "smearing" of the imaging. In the worst case, I have heard a pair of Isobariks producing nothing more than a "fat mono" sound—it was instructive to note that the listening room in that case had seating everywhere except in the traditional central position.

The sound in that room was very "musical" and I enjoyed listening to many records on that system, but it was a relief to get back in front of a pair of Quad ESLs and relax in the security of their rock-solid stereo, even if this meant that on some orchestral records the violins, for example, sounded as if they had been crammed into a telephone box to the far left of the stage, balanced only by the cellos stuffed in another overcrowded box stage-right. Proponents of the B'rik sound would say that this represents "unrealistically good imaging" and I agree, arguing, however, that it was a program fault which the Quads, with their greater stereo resolving power, had allowed me first to identify and then to decide whether it was acceptable or not.

Ignorance may be bliss, but a uniformity of imaging—everything reduced to that "fat mono"—is ultimately tedious, and I can only conclude that it is the generally low standard and unrealistic quality of recorded stereo, with its fatiguing positioning anomalies, which has led people to put the ability to present accurate stereo a long way down their list of desirable loudspeaker qualities. Perhaps Dr. Amar Bose had a point after all—the fact that a violin image is presented as being 20' wide can be forgivable when the result is ultimately enjoyable rather than fatiguing. And yet, despite all this, listening via good headphones it widely agreed to be very satisfying because it is uncannily accurate in its presentation of the lateral stereo information!

To Recap
Given that the correct positional information is present as a succession of instantaneous voltage ratios in the two information channels, then all the loudspeaker pair has to do is to reproduce those voltage ratios as sound level ratios in order—ignoring room interaction effects—that a convincing soundstage be set up between and behind those speakers. The end of your listening room becomes a wide window into the acoustic of the recorded concert hall, with every instrument its correct size and in its correct position.

Correctly Capturing Directional Information
In the first part of this article I looked at what information needs to be recorded on just two information channels in order that an accurate stereo image can be reproduced by two loudspeakers. If we agree that with "accurate stereo," instrumental directions at the original event are preserved, then each of those directions only needs to be represented by a voltage ratio between the two channels. The sum of those two ratios, integrated with respect to time, represents the stereo image, and the human brain proves to be a wonderful instrument in that it can sort out and unmix the ratios from the two signal waveforms and identify individual instruments—with their appropriate directions—within the overall texture.

This is an amazing feat, the aural equivalent of defying the 2nd Law of Thermodynamics and unmixing two differently coloured liquids after they have been poured into one container. The best that machinery can do is exemplified by a new measuring instrument from Trio (Kenwood abroad), their model 5920 Sound Image Meter. This measures what Trio term the "interaural correlation coefficient"—IACC. A dummy-head stereo microphone "listens" to the stereo soundfield and the meter examines the signals perceived by the left and right "ears." Effectively, it compensates for the time delay around the head from ear-to-ear and then compares the resultant two waveforms. If they are identical, then both channels are carrying identical information and the IACC is +1, corresponding to a monophonic signal. If the signals are identical but 180° out of phase, then the IACC is –1, a negative correlation.

The use of this Trio meter is obvious: play monophonic white-noise through your two favourite speakers in your listening room, place the dummy-head in the "stereo-seat," and the departure of IACC from "+1" tells you now precise (or rather imprecise) a stereo image those speakers will produce. The effect on the stereo of nearby room boundaries, for instance, can be immediately assessed. (With typical speakers and a typical room, Trio say that the IACC will be between 0.2 and 0.8—the nearer to 1 the better.)