Multichannel Audio Technologies

Surround Microphone Techniques

In this lecture we introduce several techniques for surround sound recording. Of particular interest to us in this lecture is the front LCR channels and accurate imaging. L-C-R Stereophonic Imaging:  The primary advantage of the center channel is for directional stability over an enlarged listening area. The second advantage concerns sound quality. It is found in a number of studies that the discrete three-channel system is preferable in comparison to the two-channel system in terms of “clarity” and “sound colour” of the center image, even when the listener sits precisely on the center line and does not move his head. It is presumed that this preference arises because the center loudspeaker is “easier” to listen to and that center phantom images principally cause some coloration and require “greater attention”.  L-C-R stereophony vs. L-R stereophony plus C In L-C-R stereophony, the center channel C offers increased directional stability of the complete L-C-R stereophonic image, which is divided into two stereophonic sub-areas. In L-R plus C stereophony the center channel C is used for a stable center image (e.g. soloist), in addition to the usual two-channel stereophonic representation of the source.

L-C-R microphone techniques derived from Stereophonic Principles: - Multiple X-Y:  One method derived from two channel stereophony is to use two coincident pairs (derived from the stereo XY technique). This is an intuitive solution, since each pair corresponds to the L-C and C-R stereophonic sub-areas. This technique has been found to work well in large orchestral situations where the large distance results in minimal crosstalk between the L and R channels. It is not effective however for smaller recording situations such as piano or drum kit recording.  

- Multiple A-B: Another configuration shown below is based on widely spaced microphones (“Multiple-A/B”). Five microphones are distributed in line across the stage width, the distance between neighboring microphones is in the range of 2m or more. Two effects are intended: Firstly, the exploitation of the precedence effect to reduce the multiple phantom sound source problem. Secondly, the provision of a “stable” phantom sound source half left between L and C and half right between C and R. Again this configuration is only useful only for large orchestral situations. It is wrong to reduce the microphone distances according to a smaller dimension of the orchestra (e.g. chamber music). The “double-main” methods or widely spaced configurations could perhaps satisfy only in a limited range of applications. So, how do we go about recording a string quartet or a solo piano?

The Three Channel Problem: When considering an L-C-R microphone technique for smaller recording areas, a common misconception is that we can easily just use three cardioid microphones associated with each channel, covering the desired sound stage, et voila…perfect stereo imaging! This assumption is wrong!!!! This will give rise to the “triple phantom sound source” which occurs due to interchannel crosstalk. In principle each 2-channel stereophonic basis C-L, C-R, L-R produces its own phantom sound source, and each of them would be located at divergent places, resulting more or less in a decrease of the localisation focus and clarity, and in coloration effects. Even hyper or super-cardioid microphones, arranged in a line do not provide enough channel separation for accurate imaging.

Ideally we want the sound source imaging to look like this:

For example, a sound source located 30° off-center right of the microphone will be perceived approximately 20° off-center right in the standard two-channel loudspeaker arrangement, due to the channel signal difference delivered from the microphone. The LCR microphone channels should translate directions accordingly, in particular show a linear curve in the center area. If this is achieved in the case of L-C-R stereophony, the desirable so called “unobtrusive center channel” will be effective. An example of undesirable imaging is shown below. This is an example of imaging from a Decca Tree (more later).

Practical Notes for L-C-R stereophony:

1. Either narrow or widely spaced microphone configurations can be used for stereophonic imaging. However, it is well-known that pure coincident microphone setups are not able to produce a satisfying natural spatial impression, due to the lack of adequate interchannel temporal relations (time-of-arrival, phase, and decorrelation).

2. Cardioid or super-cardioid microphones should be applied not only to minimize interfering crosstalk but also to attenuate the indirect lateral and rear sound and to ensure a sufficient leeway for allocating a certain portion of the indirect sound energy to the surround channels LS and RS.

Microphone Recording Angles:  A crucial quality of 2 or 3 channel main microphone is its recording angle. The recording angle is specific for each main microphone and defines the field (symmetrical to the main axis) within which sounds are received, in such a way that in the monitor room they will be localized between the two loudspeakers. An example: A sound source on the left edge of the recording angle is localized, when played back, at exactly the same place as another sound source, which is situated much further left (in the left loudspeaker). A further extension from the edge of the recording angle is not possible. Or in other words: Everything outside the recording angle is packed into one of the loudspeakers (see below). The size of the recording angle is defined by the main microphones' diverse degrees of differences of intensity and phase delay.

 

The  Recording    

   

The  reproduction  with  the  recording  angle  chosen  too  small    

In order to make sure that there are no multiple localizations it is essential that there is always only one “active” microphone pair. The two recording angles must not overlap in the middle. An arrangement of three microphones in a line would not fulfill this requirement. The three microphones must be arranged in a triangle as shown below, so that the mid-perpendicular of AB and BC are rotated by an angle of φ/4.  

 Furthermore, the directional patterns of the microphones must be such that sufficient attenuation is achieved between each AB and BC zone for accurate imaging. For small setups, omni-directional microphones are not sufficient for this. Cardioids or supercardioids would work better here, but even so, their responses have limitations as we shall now see. We will now establish a relationship between cardioid and super cardioid patterns and recording areas. This is the realm of the ‘Williams Curves’.

 

Williams Curves: It is important to note that merely knowing the desired recording angle is not enough to guarantee a successful stereophonic representation of the performance. There are in fact several physical and psychoacoustical limits imposed on the recording angle that must be considered in any stereophonic recording. The first of these relates to coincident cardioid microphones in the classic XY configuration. It has been shown by Mertens, and later by Simonsen, that for a source to be fully intensity panned left or right in a stereophonic field, the level difference between the loudspeakers must be approximately 15dB. In a coincident microphone configuration, the intensity difference between the microphones is dependent not only on the source angle, but also on the angle between the microphones, and quite often exceeds this 15dB limit. The intensity differences for a coincident cardioid microphone pair at different angular separations are plotted below against source angle. The points of interest here are where the intensity curves intersect the -15dB point (at an intensity ratio of 0.8125) as this will partially dictate the usable angle of that particular configuration. For example, when the angle between the microphones is 120o

the point at which we reach a 15dB difference occurs at a source angle of 70o. Thus, with a usable 70o recording angle for both the L and R sectors, we achieve a 140o

recording angle from a 120o

microphone separation.

       

Likewise, we can consider time difference if we look at spaced omnidirectional microphones. The relationship between source position and microphone spacing is shown below. As with the intensity difference of coincident microphones, the time difference in spaced microphones has a limit of approximately 1.1ms before 30o localisation at one loudspeaker. The recording angle can then be estimated by this psychoacoustical limit. For example, a microphone spacing of 50cm generates the 50cm curve shown in Fig.1.6. This curve intersects the 1.1mS line at a source angle of approximately 50o

(giving a usable recording angle of 100o).

A further consideration in setting the recording angle relates to directivity pattern of the microphones used. It has been shown by Williams that when one listens to a source recorded with a cardioid microphone, it is possible to detect a decrease in the direct sound as the source subtends an angle of greater than 65o

or -3dB off-axis gain. This is illustrated below. This means that as the source angle increases, the direct sound drops off in relation to the fixed reverberation level.

The consequence is a further limit on the usable recording angles possible with cardioid microphone pairs. To illustrate, consider a pair of coincident cardioids with an axial separation of 60 o. Here we expect to obtain a recording angle of +/110 o

(see 60 o curve in

intensity curves above). However, since the directivity pattern of a cardioid microphone dictates an acceptable 65 o

direct sound pick-up region, then the recording angle becomes 65 o

+ 65 o + 60 o

= 190 o or +/-95 o, as shown below.

 Thus the combination of distance and microphone angle must lead to:

Recording Angle < (Angle between microphones/2) + 65 o. Furthermore, if the microphones are to subtend an angle greater than +/-65 o

then there will exist an unacceptable decrease in the level of the direct sound to the front of the microphone pair. This represents the outer limit of microphone angle, and is applicable to spaced microphone pairs as well.  Based on these considerations Michael Williams set out a series of calculated values based on desired recording angle, microphone angles relative to each other and microphone distance, starting from coincident (0cm) to highly spaced (50cm). These are known as the Williams Curves, and are shown below.

   

   Each curve represents a recording angle that can be achieved by different combinations of the angle between microphones (y-axis) and microphone distance (x-axis). Because of the symmetry Williams defines a recording angle of e.g. 160° as ± 80°.

 Thus, if we want to use the Williams curves for 3 cardioids in the triangular arrangement shown previously, we need to define the recording angle as a combination of the two stereo zones i.e. if the angle between A and B is 60o then the total recording angle is 120o. On the Williams curves, we would look up the +/− 30 o curve for the angle between each microphone pair.  Optimized  Cardioid  Triangle  (OCT)    

The OCT array simplifies matters by only allowing the distance between the left and right microphones to change for a desired recording angle. The principle configuration of the OCT array is shown below. The preferred setup for OCT uses a forward-facing cardioid for the center channel. For the front L and R channels, two supercardioid microphones are placed at opposite ends of an imaginary line running about 8 cm behind the center microphone. These two microphones should be 40 – 90 cm apart, depending on the required recording angle, and must face squarely outward, away from center.

 Good separation between the LC and CR sectors is obtained with this method. For example, sound originating from half right is picked up only very weakly by the left microphone. Sound from the extreme right will be picked up directly on-axis by the right-facing supercardioid (0 dB) and by the forward-facing cardioid (attenuated by 6 dB due to its directional pattern). Finally it will be picked up, with a delay caused by the increased distance, on the rear lobe of the left-facing supercardioid. The polar pattern attenuation for this will be 10 dB and the polarity will be inverted. In the case of frontal sound directions (Ω = 0°) the following interchannel level differences ΔL result:

Source

These level differences are based on the directivity pattern of unidirectional microphones:

Distance b depends on the recording angle as shown below. The recording angle is generally between 90 – 160°. Distance h = 8 cm. If cardioids are used instead of super-cardioids, h = 12 cm.

 Array Calculation Tool: A calculation tool for triangular arrays, called “Image Assistant" has been developed by Helmut Wittek and can be found at www.hauptmikrofon.de. It offers the calculation of localisation curves (and corresponding recording angle), interchannel level and time differences, overall sound level, for any two-channel or three-channel microphone configuration, including omni, cardioid, super cardioid, broad cardioid, figure of eight characteristics.

Breaking the Rules: Decca Tree

Spaced omnidirectional microphones have a nice, open spacious sound. The wider the separation, the more spacious, but if they get too wide, then you get a ‘hole-in-the-middle’ effect. In the Decca tree, three omni-directional microphones are widely spaced in a triangle configuration. Because of the wide microphone spacing a recording angle does not exist and it is not suitable for accurate stereophonic directional imaging. It is used to produce an open, spacious sound, combined with a solid central image. The Decca Tree was originally conceived by the recording engineers at English Decca Records. It utilized three omnidirectional microphones situated at the ends of a large T-shaped fixture. The spacing between the left and right microphones is approximately 2 meters, and the central microphone was in front of these by about 1.5 meters. Placement of the array is generally a few feet behind and about eight to ten feet above the conductor’s head. As mentioned earlier, in most 3/2-stereo recording situations it is advantageous to use uni-directional microphones to reduce the energy of indirect sound and to provide headroom for allocating the indirect sound energy to the surround channels. For this reason it may be useful here to replace the omnis of the Decca-Tree by cardioids, each of them facing the front (off-center angles = 0°). This does not change the directional characteristics of the tree, but the indirect sound level is lower. A similar triangle configuration is applied in the Fukada-Tree with cardioid microphones used.

Decca Tree using Neumann M50s

Typical Decca Tree Setup Over Conductor