This Is Stereo! PART 5

Building on the concepts explored in previous parts of this series, it’s time to weigh up some of the pros and cons of coincident and spaced arrays.

Previously in this series I focused on a range of coincident microphone techniques, and you’ll recall that with all such techniques, perceived position on playback depends entirely on level difference in the two channels. This time, we turn our attention to spaced stereo arrays, which introduce time‑of‑arrival differences between the two microphones. These give an entirely different sense of stereo imaging — less precise and more diffuse — but it’s often also subjectively appealing, and there can be benefits as I’ll explain below.

Precision Position

First, let’s pick up on that point about ‘precision’. With coincident arrays, equal volumes in each channel create a ‘phantom’ central image source in the speakers, and introducing a level offset between the channels moves the sound source towards the louder channel’s speaker. It’s not equal for all frequencies, though.

For mid and high frequencies, which tend to be portrayed wider than intended when auditioned over stereo loudspeakers, an offset of around 4dB places the sound about a quarter of the way out towards one speaker. 8dB is half‑way, 12dB is three‑quarters, and 16dB puts the source firmly in one speaker (see Diagram 1).

Diagram 1: Perceived pan position caused by differences in inter‑aural level (left) and timing (right).

These are only rough guide numbers — different listeners and different monitoring environments might need smaller or larger offsets to achieve the same perceived image positioning. But slightly larger offsets are typically needed to match positions for the lower frequencies. It’s why EMI developed the Stereosonic Shuffling correction process I described back in Part 3. (There are other issues that affect the precision of stereo imaging over speakers, and various remedies, but I’ll save that for another article!)

Coincident mic arrays aren’t the only means of creating the impression of different spatial positions, though. As Dr Fletcher discovered with his ‘curtain of sound’ experiments, reproducing small inter‑channel timing differences also works — though arguably with less well‑defined image locations.

Remember, the ear/brain uses inter‑aural timing differences to help locate sound sources around us, and for a source located 90 degrees to one side, the difference in arrival times for a sound at each ear is about 0.67 milliseconds. Yet, if we introduce that kind of time delay between channels replayed over stereo loudspeakers, the sound source will be perceived as being barely half‑way towards the earlier loudspeaker — this is because sound is coming from both loudspeakers, each introducing their own time delays for sound arriving at each ear. Essentially, it’s an unnatural and confusing situation, as far as the ears/brain are concerned. We can still create the impression of a sound source being fully in one loudspeaker, but the time offset between channels actually needs to be around 1.5ms (again, see Diagram 1).

So, if we’re to utilise inter‑channel timing differences to create the illusion of sound source positions between stereo loudspeakers, how can we arrange a pair of microphones to capture only suitable inter‑channel timing differences that relate proportionally to the real sound source positions? It’s probably fairly obvious that to capture time‑of‑arrival differences between the two channels, we’ll need to space the mics apart by a small distance: sound from a central source will arrive at both mics at the same time (giving a central phantom sound image), but sound from a source located to the left will reach the left microphone slightly before the right one, resulting in a small time difference between the channels.

With the two microphones spaced apart, the inverse‑square law of sound propagation — sound energy decreases with distance from the source — means that the more distant mic will also always receive a slightly lower sound level than the closer mic, so a small level difference between channels will be captured too. However, if the mics are reasonably far away from the sources, but not too far apart from each other, any channel level difference will be insignificant in terms of sound image location information. Moreover, if we use omnidirectional microphones, the relative angle of sound incidence won’t affect the sound level either...

Spaced Omnis

So, we have a pair of omnidirectional microphones spaced apart specifically to capture time‑of‑arrival differences — a technique generally referred to as ‘spaced omnis’ or A‑B stereo (in contrast to the coincident or X‑Y stereo technique). As an aside, it’s worth noting that Alan Blumlein also experimented with spaced omnidirectional mics in the 1930s, coming up with an arrangement most know today as the Jecklin disc; I’ll discuss that concept another time.

If two omnidirectional mics are mounted coincidentally, there can be no level or timing differences between the signals they capture. Whatever the sound source’s position relative to the mics, identical signals in both channels will create a phantom central image over the stereo loudspeakers. If the mics are gradually moved apart, though, sound from a source located to one side of centre‑stage will reach the closer mic slightly before the more distant mic, resulting in a small inter‑channel timing difference.

So, what kind of mic spacing do we need to generate an inter‑channel timing difference sufficient to place the sound fully wide in one loudspeaker? Consider a sound source, located at 90 degrees to the microphones (ie. fully left), that we want to perceive fully in the left‑hand speaker. As suggested above, the sound waves must arrive at the left‑hand microphone 1.5ms before passing across the right‑hand microphone — sound travels around 34cm per millisecond, so achieving that delay requires the mics to be spaced 51cm (34 x 1.5) apart, as in Diagram 2.

Diagram 2: If the distance between mics is less than 51ms, no sound will appear to be ‘fully panned’.

Put the mic capsules closer together than 51cm, then, and it will always be impossible to generate a timing difference between channels as large as 1.5ms, whatever the source position — the image width on loudspeakers will always be narrowed, and collapses to a phantom centre as the capsules approach one another. Conversely, if the capsules are moved further apart, a sound source would need to move a smaller distance from the centre line to generate the 1.5ms difference.

This gives us a very practical way of controlling the Stereo Recording Angle (SRA) from spaced‑omni microphones: increasing the capsule spacing reduces the SRA. We’ve seen that a capsule spacing of 51cm gives an SRA of 180 degrees, while increasing the spacing to 73cm reduces the SRA to about 90 degrees. Inevitably, if the capsules are moved too far apart, the channel timing differences generated by off‑centre sound sources can exceed 1.5ms. That breaks the stereo system — instead of a continuous spread of sound between the two loudspeakers, we perceive ‘puddles’ of sound localised at each speaker, with a gaping ‘hole’ in the middle of the sound stage. This ‘hole‑in‑the‑middle’ effect was quite common in early commercial stereo recordings, partly by design, as record labels tried to emphasise the benefits of two‑channel recordings: “Completely different sounds from your two loudspeakers — so much better than mono!” Some labels cured this hole‑in‑the‑middle problem by adding a third, central, omni mic... Again, I’ll explore that another time.

SRA Rules

In Part 3, I defined a Rule 1, for fine‑tuning the SRA of a coincident array by adjusting the mutual angle. We can now add a second rule for the capsule spacing in spaced arrays:

• Rule 1: Reducing the Mutual Angle (MA) increases the Stereo Recording Angle (SRA).
• Rule 2: Reducing the Capsule Spacing (CS) increases the...