Features

Multi-Mono Stereo Recording
But, to get back to the theme of this article, how can these voltage ratios be derived? The most basic way is to take a monophonic signal and assign a voltage ratio to it by "panning" it to a position with a "panpot." This gives an image with unambiguous positioning and this panpotted multi-mono image is to be heard on any rock recording. Though it doesn't have positioning anomalies, it is no more than a slightly more complicated version of the notorious "now it's here—now it's there" ping-pong recordings of the late '50s. Despite effects such as artificial reverb (also panpotted to a suitable place in the stereo stage) and anti-phase gizmos such as Aphex, this unsophisticated approach to stereo recording doesn't recreate an image of a real event but gives a montage of mono sound-sources. Perhaps appropriately, considering the cultural position of rock music, it resembles a poster rather than a photograph, and any resemblance to an original event is almost unnecessary.

Binaural Recording
Classical music, however, has evolved so that the individual musicians within any kind of ensemble achieve a satisfactory internal balance, and anyone listening will perceive a real continuous image with both width and depth. There should be, therefore, some way of recording that image with its instrumental directions preserved so that they will be reproduced accurately.

The obvious way would be to use a pair of omnidirectional mikes where the listener's ears would be, intercepting the soundwave entering each ear so that when played back over headphones, that soundwave is recreated, hopefully exactly. The brain then has the same information to process as if the listener had been at the original event. However, the mandatory use of headphones for playback has meant that this binaural technique, despite its uncanny success in recreating the soundfield, has never caught on.

Blumlein Stereo
Blumlein's genius lay in his working-out of a mike technique that automatically encoded directions as amplitude ratios between two channels. His basic idea was to use two figure-of-eight microphones with their axes at right-angles (fig.4). They must be coincident in space, an impossible condition but one that can be arranged for lateral imaging—which is all we're concerned with—by placing one just above the other. Ralph West in his article "Blumlein Stereo" (HFN/RR June 1979) described graphically and succinctly the principle involved, but it is worth repeating.

The response of a figure-of-eight mike is proportional to cosine Θ, where Θ is the angle the sound-source makes with the mike axis. If the crossed-pair is positioned so that sound-sources to the far left and right are on the left- and right-hand mike axes (Ml, Mr) respectively, because the mike axes are at 90° a source on Ml will produce a maximum level (cos Θ = cos 90° = 0) into the right-hand mike. As the source moves on an arc from left to right, the left-hand mike signal reduces to zero while the right-hand mike signal rises to a maximum. At any intermediate position, the ratio of the amplitudes between the two mikes is cos Θ:cos (90°–Θ) or cos Θ:sin Θ, an effectively linear relationship.

If you listen to Blumlein's original experimental recordings made with crossed figure-of-eights, as he walks from far left to right, his image moves steadily across the stereo stage. As he moves away from the mike, his image remains narrow but is unambiguously placed further away as the direct/reverberant ratio is correctly captured. This, in fact, is the great advantage of any coincident technique: as well as capturing the directions of the direct sound-sources accurately, every reflected wavefront making up the reverberant field is also correctly placed. The ambience is coherent and the reproduced lateral stereo stage is refreshingly free of anomalies—instruments are convincingly placed within the room in which there recording was made.

And yet many listeners can be upset by this. In a challenging article in the January 1977 issue of Studio Sound ("The echo of Dashion") Peter Fellgett commented on this, saying that "some professional hear as "unclear" or "fuzzy" any recording which provides a definite sense of the acoustic space in which the performance took place...they see sic anything which contributes specific information about the recording space and its acoustic ambience as an intrusion which they interpret as confusing."

So strong has this feeling been amongst record makers that most classical recordings, while possessing their fair share of reverberation, present the stereo stage within an anomalous amorphous ambience. Critics and consumers, grown used to this sound, now react adversely to the more natural recording, reinforcing the trend away from capturing coherent ambient information: Peter Fellgett, again—"Broadcasts or recordings that sound really correct are so rare that reviewers tend to shy away from them as not fulfilling their expectation of "hi-fi" reproduction."

It must be remembered, however, that record producers and engineers are intelligent people and that this gradual move away from naturalistic recording isn't due to perversity alone. Look again at fig.4: it will be obvious that not only will the orchestra and its ambience ahead of the mikes be captured, but so will all the ambience behind and to the sides. The result it that unless the engineer is experienced and knows the best place to put the mikes in the chosen hall, far too much reverberation will be captured and the recording, though realistic, will be too "wet"—a sound that reminds one critic of sitting in an empty school hall and wondering why the audience hadn't turned up.

And even if the engineer has managed to find that "sweet spot" for his crossed figure-of-eights, with the optimum direct/reverberant ratio, they might then be far too close to the performers, unacceptably exaggerating the width of the stereo stage. There is also a problem in that so far I have discussed only "perfect" microphones. The response patterns of real figure-of-eight mikes tend to narrow at high frequencies, so central images would tend to be a bit down in top: this gives rise to a discernible "hole" in the centre of the stage (the degree of which is dependent on the mike) which many listeners find unacceptable.

Other Coincident Techniques
An alternative approach would still be to use a coincident pair, but employing microphones with a suppressed rear response, such as cardioids (fig.5), so as to reduce drastically the contribution from the reverberant field. If these have their axes at 90° and have the perfect cardioid response of output proportional to 1 + cos Θ, then a soundsource on the Ml axis will produce a maximum level (1 + cos Θ = 1 + cos 0° = 2) into the left-hand mike. However, whereas the crossed figure-of-eights give a null—zero output—into the right-hand mike, a soundsource on the Ml axis will still give a relative output (1 + cos Θ = 1 + cos 90° = 1) into the right-hand cardioids. The ratio of amplitude is thus 2:1, a difference of only 6dB!

Crossed cardioids at 90°, therefore, give an inherently narrow stage, but in practice the situation is even worse than this. "Real," as opposed to "perfect," cardioid mikes have difficulty in maintaining their pattern at low frequencies, becoming more omni in character; bass frequencies, therefore, are effectively recorded in mono by 90° cardioids and it has been remarked before that with such an arrangement the LF ambience seems to "pool" in the centre. Their pattern also tends to narrow at high frequencies, again giving rise to treble "hole in the middle."

Opening out the angle between the cardioids' axis gives an increased amplitude difference between the two channels for extreme left-hand sound sources: an included angle of 135°, for instance, will give a difference of around 10dB, so the stereo stage will be significantly wider than that produced by a 90° pair. Placing the cardioids back-to-back ie a 180° angle, will give the maximum channel separation, but also gives maximum sensitivity to far left and right. Thus reverberation will be emphasised at these position, giving a "pulling" of the ambience to the sides. Central images will be 90° off-axis to both mikes where the HF response of "real" cardioids doesn't hold up, giving a treble hole in the middle. Instruments that increase in treble as they are played louder will, if central when quiet, "splash" to the sides with 180° cardioids.

If a 90° crossed-pair of mikes with a hypercardioid pattern is used, this will give a wider stage due to the reduced sensitivity to off-axis sound sources. Many professionals do prefer to use crossed hypercardioids because, in addition to their wider stage, the fact that their pattern does possess a slight antiphase rear lobe gives extra "space" to recorded reverberation.

There is another way of widening the stage from a crossed-pair of directional mikes, if use of more fundamentally correct crossed bidirectional mikes is unpractical. This is to take advantage of the well-known Haas or "precedence" effect, best known from its effect on off-centre listening to a pair of loudspeakers. In that situation, if the same signal is fed to the two loudspeakers, introducing a time delay in the feed to one of the speakers will shift the otherwise centrally placed image for a centrally placed listener towards the other speaker. If that delay is roughly 2ms or more, then the image will be displaced towards one side. What, then, if we take our coincident cardioids and introduce a time delay for off-centre sources by spacing them slightly?

The additional time delay information will reinforce the amplitude-only information with its limited separation and, dependent on the order of the maximum time delays, could well widen the narrow amplitude-only stereo stage to an acceptable degree. Many recording engineers who have given up on figure-of-eight microphones therefore use crossed cardioid or hypercardioid mikes separated horizontally by a distance which they feel gives the necessary compensation. ORTF, the French broadcasting system, for instance, favour cardioids 17cm apart with an included angle of 110° (fig.6).

While this approach departs from the precise, mathematically correct approach of Blumlein, and is less elegant in that it represents the arrangement of two "wrongs" so that they cancel, it is of undoubted practical value. However, the dilution of Blumlein's thinking has led to the situation where confused recordists assume that any coincident pair of mikes must be intrinsically "correct," and to the widespread adoption of what almost seems to be "arbitrary arrangements of semi-coincident, semi-spaced arrays using arbitrary directional patterns and inter-microphone angles" as Peter Fellgett put it.