Last month we investigated the various coincident techniques for stereo recording developed by Alan Blumlein in the early 1930s; this time we'll be covering some alternative techniques using spaced microphone arrays.
Some of the earliest stereophonic experiments were made in America under the direction of Dr Harvey Fletcher at Bell Laboratories, as mentioned in the first part of this feature. One of the techniques investigated was the 'Wall of Sound', which used an enormous array of microphones hung in a line across the front of an orchestra. Up to 80 microphones were used, and each fed a corresponding loudspeaker, placed in an identical position, in a separate listening room.
The operating principle was that the array of microphones 'sampled' the wave-fronts of sound emanating from the orchestra, and these exact wave-fronts were recreated by the loudspeakers in the listening room. The results were extremely good, with remarkably precise imaging and very realistic perspectives. However, the technology of the '30s was such that recording or transmitting 80 discrete signals was simply not practical.
Consequently, the initial microphone array was systematically simplified to find the minimum number of microphones that produced acceptable results. The general consensus was that three microphones and three loudspeakers represented the best compromise between high-quality imaging and practicality.
Today, the three-spaced-microphone technique is still in widespread use (one form being the Decca Tree) and the three-loudspeaker arrangement is the standard method of frontal sound reproduction in every cinema!
So, what is the disadvantage of spaced microphone techniques compared with the coincident systems? Well, the main problem has to be mono compatibility. Any array that has multiple microphones spaced apart from each other will capture the sound from a given source at different times. If the outputs from all of the microphones are mixed together (to produce a single mono signal), the sound will become coloured because of a process known as 'comb filtering' -- the beginnings of phasing or flanging. In a severe case, the comb filtering may alter the sound to such an extent that an orchestra will sound as if it's at the other end of a long cardboard tube. The greater the number of combined microphones, the worse the effect is likely to be.
However, if you can guarantee that recordings produced with a spaced microphone array will not be combined to make a mono signal, the comb-filtering problem becomes totally irrelevant -- the argument adopted by many of the organisations that record classical music.
Since virtually all of the classical music catalogue is on CD or cassette these days, mono replay is no longer an important consideration -- when was the last time you saw a mono CD or cassette player? Whether you're listening on a serious hi-fi, in the car, on a mini-system, or through a 'Brixton briefcase', it will almost always be in stereo. Even broadcast music is in stereo on Classic FM or Radio 3 for the vast majority of listeners.
So classical music, in particular, is often recorded with spaced microphone arrays because mono compatibility is not an issue -- but why would anyone want to use spaced arrays? What is wrong with coincident systems which offer mono compatibility as standard?
As we saw last month, all coincident systems have to use directional microphones in order to create the necessary level differences between the two channels of the stereo system. Directional microphones rely on the pressure-gradient principle, which has an inherent problem with low frequencies. The mechanical design of the microphone diaphragm assembly has to compensate for the inadequacies of the pressure gradient by making the diaphragm resonate at very low frequencies. Although this can achieve an acceptable frequency response, it generally compromises the sound quality, restricting the smoothness and extension of the very lowest frequencies.
On the other hand, omnidirectional microphones do not suffer from any of these compromises. They have very smooth and extended low-frequency regions, with very even off-axis responses, both of which are very desirable characteristics. The only problem, of course, is that omnidirectional microphones do not work terribly well as coincident pairs because they do not produce level differences proportional to the angle of incident sound. Omnidirectional microphones can only be used to record in stereo if you space them apart and deliberately record timing differences.
I mentioned last month that reproducing the kind of timing differences captured by a spaced microphone array over a pair of loudspeakers would confuse our hearing system and therefore not produce good stereo images. This was only a half-truth, I'm afraid! It is true that replaying a stereo recording with timing differences between the two channels leads to a confusing set of time-of-arrival differences for our ears, but the sound is normally still perceived as having width and a certain amount of imaging information, and it usually sounds a lot more spacious than a coincident recording.
The problem (as far as Alan Blumlein was concerned, anyway) is that, apart from the mono compatibility issues, the imaging is not very precise and often seems to huddle around the loudspeakers rather than spreading uniformly between them. In really bad cases, the recording may even appear to have a hole in the middle!
If I were to compare the two main types of stereo recording as if they were paintings (a ludicrous thing to do, but I'm going to anyway!), then good coincident recordings are like etchings or line drawings -- very precise imaging, lots of detail, leaving nothing to the imagination. On the other hand, spaced microphone recordings are more like water colours -- the detail is blurred, and the essence is more about impression than reality.
Many people specifically prefer the stereo presentation of spaced-pair recordings, finding them easier to listen to than coincident recordings. There's nothing wrong in that -- as far as the recording engineer is concerned, this is just another technique with a collection of advantages and disadvantages over the alternative formats. It's up to you which technique you use, and as long as you are aware of the characteristics of each system, you are in a position to choose wisely and should be able to achieve the sound quality you seek very quickly.
The simplest spaced-microphone technique is to place an identical pair of omnidirectional mics a distance apart in front of the sound source; most engineers would generally choose a spacing of between a half and a third of the width of the actual sound stage. For example, if you set out to record an orchestra, typical positions might be a quarter of the way left and right, either side of the centre line. The distance between orchestra and microphones will depend on the acoustics of the environment and the kind of perspective you want to achieve. For the recording, each microphone feeds its corresponding track on the stereo machine.
The potential problem with this arrangement is a hole in the middle of the stereo representation. The simplest way to avoid this disastrous situation is to bring the mics closer together, but this will affect the spaciousness of the recording, the whole thing tending to become rather narrow and lifeless. The optimal position is often a little harder to find than might be imagined at first.
Although most people use the spaced technique purely so that they can take advantage of the qualities inherent in omnidirectional microphones, there is no reason why you should not use directional microphones in a spaced array -- a very well-known classical music recording engineer, Tony Faulkner, often uses figure-of-eight microphones, for example. The advantage of using directional mics is that it is possible to reject some unwanted signals (typically reverberation) while retaining most of the other advantages of spaced-mic recordings.
Other spaced techniques that use directional microphones include the ORTF format and the NOS system (both named after the European broadcasting companies that developed them). These are often called 'near coincident' techniques because they combine the level difference recording characteristics of directional coincident microphones, with spaced arrays.
In the case of the ORTF technique, the basic configuration uses a pair of cardioid microphones with a mutual angle of 110°, spaced about 17cm apart. The NOS variant has a 90° mutual angle and a spacing of about 30cm, and -- just for the record -- the Faulkner array uses a pair of figure-of-eights, both facing directly forward, but spaced by about 20cm.
I often use the ORTF or NOS techniques, generally with good results, and I recommend that you experiment around these basic arrangements to find out what works best for you.
Binaural recording is one of those techniques that seem to have a cyclical life. It becomes very popular for a while, then seems to disappear without trace, only to be re-invented a few years later...
Binaural recording is a basic two-microphone spaced-pair technique, but it is rather specialised in that it only works effectively when listened to through headphones. The principle is to replicate the way our ears capture sounds, and replay those sounds directly into the corresponding ears.
Our ears have a hemispherical polar pattern, largely dictated by the lump of meat cunningly positioned between them. As we saw last month, the head creates sound shadows and timing differences for the two ears, so a binaural recording format has to replicate those actions.
The easiest technique is simply to clip a couple of small omnidirectional microphones (tie-clip mics are perfect) to the ears of a willing victim (the arms of a pair of glasses would be less painful!). If recording an orchestra or band, the human mic stand would have to be persuaded not to move their head, but stunning results can be obtained if the microphones are recorded on a cassette or DAT Walkman as you go about your daily chores. Crossing a busy road can cause very entertaining reactions in the listener -- and see what happens if the recording includes your morning ablutions: listening to the binaural sound of someone brushing their teeth is an experience in itself!
A rather more practical method is to use a Jecklin Disc (also known as a 'Henry' in some circles), which mimics the fundamental acoustic aspects of the average head. The disc can be made from perspex or plywood, typically about 25-30cm in diameter, with a mounting point for the microphone stand on one edge, and fixings for a pair of microphones arranged through its centre.
The surface of the disc should be covered in some kind of absorbent padding to avoid reflections from the disc surface back into the microphones, and the mic capsules should be mounted about 15-18cm apart on opposite sides of the disc.
The operating concept is that the microphone-spacing matches that of our ears, and the disc provides the sound-shadowing effects of the head; thus the whole technique should be able to capture signals in the microphones which will closely match those of our own ears. When replayed over headphones, the signals from the disc mics are fed directly into our ear canals, bypassing the effects of our own head- and ear-spacing -- effectively transporting our ears directly to the recording venue.
In practice, the results vary from being incredibly three-dimensional and realistic, to forming a stable and solid image behind, but not in front of, the listener's head. The differences are probably due to the difficulty of accurately matching the dimensions of the recording system to the listener's own physique. If you really want to go overboard, the state-of-the-art binaural technique uses a fully bio-accurate dummy head (a common alternative name for the binaural system is 'dummy-head recording'). Several manufacturers produce anatomically accurate heads, often with the complete torso. Even greater accuracy can be achieved by using carefully shaped pinnae around the microphone capsules and even replicating the various mouth, nose and sinus cavities within the human head!
Interestingly, binaural recordings replayed over loudspeakers manage to convey a sense of stereo width and movement without having any accurate imaging qualities. This facet of the technique is often used to advantage in the production of sound effects for radio and television. In general, sound effects -- especially atmospheric effects -- should convey the environment, but must not distract from the foreground dialogue or action.
Imagine a scene from a radio drama where two actors appear to be conversing on a busy town street. If coincidentally recorded effects were used, the sounds of footsteps and buses travelling across the sound stage could be intensely distracting. However, a binaural recording of the same atmosphere, while being very realistic over headphones, is far less distracting over loudspeakers. Scale, width, perspectives and movement are all conveyed to the listener, but in a laid-back manner that is often far more effective.
Spaced techniques allow the engineer to take advantage of the inherent quality of omnidirectional microphones in stereo recordings, particularly their extended and even low-frequency response, and their smooth off-axis pick up. The only problem to be aware of is the potential for comb-filtering effects when spaced microphones are combined, and the more the mics, the worse the effect is likely to be, although it is very hard to predict the audible results.
There are no rules about spacing microphones -- it really is a case of trying an arrangement and listening carefully to the results, then moving things about until you find what you are after. Thinking about the physics of the whole thing helps but, in practice, placing spaced microphones is really a black art, and the best results are almost always obtained by trial and error.
Binaural techniques are a lot of fun, and can be stunningly realistic, but most people prefer to listen over loudspeakers rather than headphones, so the technique is of limited practicality.
Most of my own best recordings have used spaced systems to overcome the weaknesses of coincident techniques (generally, excessive precision and a lack of spaciousness), and my personal favourite arrangement is the near-coincident ORTF setup. This rarely fails to provide the kind of sound I like, and is usually a good starting point from which to build the final sound.
I have used variations on this theme to form the basis of recordings of everything from solo acoustic guitars, complete drum kits, Leslie speakers on Hammonds, and live jazz bands in clubs, up to full concert orchestras.
However, your likes and dislikes, in terms of stereo sound stages and imaging details, are bound to be different from mine, so don't take my word for it, go out and experiment -- it really is the only way!
The most commonly used spaced-pair technique is probably the Decca Tree. This was developed many years ago (some time in the early '50s, in fact) to allow the use of omnidirectional microphones to record in stereo.
The basic arrangement is to mount three microphones in a triangular pattern, the central microphone being forward of the others. Dimensions are not particularly critical but, typically, the two rear microphones are about 140cm apart, with the central microphone about 75cm forward of them. The exact dimensions may be varied to suit the size of the sound stage being recorded, and, depending on the polar accuracy of the omnidirectional microphones used, it may help to angle the outer microphones towards the edges of the sound stage so that the microphones' best high-frequency response favours these parts.
The left microphone is recorded on the left channel of the stereo recording, and the right mic on the right channel, as you would expect; the central microphone is distributed equally between the two channels. Although combining the central microphone with the two edges is potentially risky in terms of comb-filtering effects, the hazards are far outweighed by the advantage of a very stable central portion of the sound stage, avoiding any possibility of a 'hole in the middle'.
This extra stability is not only due to the mere presence of an additional microphone covering the centre of the sound stage, but also because the central microphone is forward of the others. The slightly closer proximity to the sound stage means that it will capture sounds before they arrive at either of the other microphones. On replay, this will cause the sound stage to build from the centre, expanding to the edges as the out-rigger microphones capture the sound fractionally later. It is a very subtle effect -- one that works at a subliminal level -- but is crucial to the effectiveness of the Decca Tree format.
In most cases, one primary technique rarely produces the results we want in our recordings. Last month we saw how combining a coincident pair with spot microphones was an effective technique. A similar combination of spaced mics and spot mics also works well, but many engineers prefer to add a spaced array to the full coincident/spot combination. This adds more spaciousness and ambience to the recording and is simply achieved by placing a pair of omnidirectional out-riggers towards the left and right edges of the sound stage. The effect of the omnis is to provide a much richer and more substantial sound, while the coincident pair provides most of the imaging accuracy and the spot mics highlight the inner detail and lift the weaker instruments.
The idea can be used on a drum kit, where spot mics are placed close to each drum head to capture the slap and attack of the sticks on the skins, and a pair of omnidirectional microphones is placed some distance away to give a broad and spacious stereo image. If the microphones are placed low down, towards the floor, they will tend to favour the drums rather than the cymbals, and if they are higher up the reverse is true.