Articles and whitepapers

Whitepaper: Surround with Fewer Speakers (1/3/2007)

By Mark Dodd and Jack Oclee-Brown, KEF Audio (UK)

In recent years the experience of watching a movie at home has been transformed from 'watching the TV' to 'experiencing home cinema' by the use of multichannel sound to envelope the listener and provide a more absorbing experience. Achieving this effect requires more audio channels and more loudspeakers. While in a commercial theatre this is not a problem, in consumer audio the extra wires and loudspeakers may be difficult to reconcile with the domestic environment. It has become desirable to deal with this conflict by reducing the number of loudspeakers while still attempting to convey as much of the additional channel information as possible.

This paper reviews this concept and explores methods by which this concept has been achieved.

INTRODUCTION

Multichannel audio is now widely available to the consumer. These multichannel formats demand numerous loudspeakers. While five or more loudspeakers in the room may be acceptable to the enthusiast, to the normal consumer this is an undesirable compromise. A more widely acceptable compromise, however, is likely to be a small reduction in the surround performance in exchange for fewer 'boxes' around the room.

In order to realise a virtual surround system it is necessary to devise some means of tricking the human auditory system into thinking that a sound is coming from somewhere that it is not. Our aim is to create "virtual soundfields" which envelope the listener while only utilising few loudspeakers.

Perception

Comparison of the signals at each ear allows the brain to determine the spatial characteristics of the soundfield in which the subject is located. If the two signals are sufficiently coherent, sources may be localised without too much difficulty. In this context, the degree of coherence is a measure of how perfectly these signals can destructively interfere. The coherence of two signals follows from how well correlated the two signals are [1].

A conventional loudspeaker is a very coherent radiator of sound. It is relatively easy for the auditory system to determine the direction from which a loudspeaker radiates. With multiple loudspeakers the location of 'virtual sources' produced by amplitude panning is similarly clear [2].

Where the sound at each ear is incoherent the sound tends to be attributed by our cognitive processes to the environment. This leads to the perception of 'spaciousness' that embodies a sense of awareness of the surrounding space [3].

Multiple Loudspeakers, together with appropriate signals, are able to produce spatially incoherent sound-fields. It is common for multiple loudspeakers to be used for the rear channels in cinema audio systems in order to give a more spacious surround effect. Spaciousness can also be experienced in a similar way, albeit to a lesser degree, with stereo.

Localisation

An important feature of our anatomy is that we have two ears spaced some distance apart. The physical arrangement of the ears, on either side of the head, along with the anatomy of the torso, head and pinnae, provide a difference in the two signals arriving at each ear.

The difference in distance between source and ears contributes an azimuthally level difference and arrival time for incident sound. The relative difference in level & arrival time between the ears is commonly referred to as the inter-aural level difference (ILD) and the inter-aural time difference (ITD). These two effects, ITD & ILD, are at the root of the mechanism that allows us to determine the direction of coherent sounds.

One effect of the full pinnae, head and torso geometry is to introduce direction varying frequency response differences. This information is also used during localisation. It is believed that these differences, together with head movement, allow the perception of the height of a source.

The full transfer function between a source in the far field and the ear canal is referred to as the Head Related Transfer Function (HRTF). Due to the variation of shape and size of the pinnea, head and ear canal the HRTF varies between individuals.

Spaciousness

A diffuse field has the property of random incidence. Acoustic waves travel in every direction with equal probability. In this situation we could expect the soundfield at our ears to be totally uncorrelated and the localisation mechanisms discussed above cannot function.

To illustrate this consider the situation shown in figure 1 where the subject is surrounded by an applauding crowd. In circumstances such as these the signals arriving at the ears are uncorrelated. The normal spatial mechanism based upon signal arrival times and signal level differences cannot function. Instead of a perception of localisation a perception of spaciousness is experienced. The subject is aware that the soundfield is around them, they are enveloped by it, but they are unable to determine the locations of the individual sources.

Figure 1 - the subject is surrounded by an applauding crowd

This sensation is often experienced in a large reverberant space, particularly with continuous sounds. Localisation becomes difficult and is replaced by a sense of spaciousness. In such a space one is much more likely to be in a diffuse field, particularly if the source is distant and continuous.

In structural acoustics the inter-aural correlation coefficient [4] is one commonly used metric for the degree of diffuseness of a reverberant field. A low IACC is desirable for concert halls since it indicates a diffuse sound which will be percieved as spacious

SURROUND QUALITY OBJECTIVES

Surround sound signals may include ambient and discrete sounds. Holman writes: "surround sound is used best aesthetically where is does not contain many discrete sounds that draw attention to themselves, for there is no supporting picture as there is with the screen channels and one's attention is drawn to the loudspeaker"[5].

This statement is interesting since it implies a criterion for performance for surround reproduction that is more related to listener envelopment than precise localisation of discrete images all around the listener. In fact the localisation of a surround speaker itself is regarded by Holman as a failure to maintain the desired illuson. The front sound has a totally different criterion, precise imaging being a key performance attribute.

Although the technical achievement of delivering localised images around the listener is unquestionable we should not forget that the purpose of surround sound is to enhance the illusion provided by cinematic material. It is the opinion of the authors that this is best achieved by enveloping the listener in a diffuse field.

CREATING VIRTUAL SOUNDFIELDS

As stated in the introduction, our aim is to create "virtual soundfields" which envelope the listener while only utilising few loudspeakers.

A Virtual surround system must trick the auditory mechanism into believing that a sound is emanating from a location where it is not. There are several ways that this has been attempted as described in the following sections.

Application of Head Related Transfer Functions

HRTF measurements contain all of the information necessary to deduce the approximate signals at the ear canal corresponding to a source anywhere in the soundstage. Using HRTF measurements it is simple to convert a multichannel signal into a two channel signal; one containing the signal to be delivered to the left ear, one containing the signal to be delivered to the right ear.

These signals must then be delivered to the ears for the listener to perceive the correct image. The delivery of the signals is difficult. Playing the signals through conventional loudspeakers is not effective since the signals are not delivered exclusively to each ear. From the left ear the listener is able to hear the right speaker and vice-versa. This effect is known as crosstalk.

The most direct way to reproduce the signals at the ears without crosstalk is to use headphones. The HRTF approach is often the basis of virtual surround systems using headphones [6]. However, delivery with headphones is not ideal, the experience is solitary and the listener is isolated from the actual acoustic environment in which they are located. There is also some imaging performance degradation, when locating a sound a subject will subconsciously adjust their head position modifying the HRTF in order to refine the location. With headphones it is clear that moving the head will have no effect on the signals at the ears, this can result in a "sound in the head" sensation.

A more effective way of delivering the signals is by use of loudspeakers and a crosstalk cancellation system. Crosstalk cancellation requires a network of filters which are the inverse of the transfer function between loudspeaker and listeners ears. This allows the signals arriving at the entrance to the listener's ear canals to be controlled. As this method is based on system inversion, problems can occur if the acoustic paths are badly conditioned [7]. The Optimal Source Distribution (OSD) [8] approach uses a speaker array designed to aid the filter inversion. When used with binaural recordings this system is remarkably convincing.

More commonly, systems have been designed to work with normal loudspeakers [7]. In both cases the approach used relies on the listener being located exactly between correctly positioned speakers and close enough for the direct sound to be louder than the listening rooms' reverberant sound.

Application of Local Boundary Reflections

A direct method of producing an image of a source in a different location to the loudspeaker is to use reflection. When a sound is reflected from a smooth large flat surface an image source is produced at a location determined by the angle of incidence.

To produce a virtual source by reflection the sound must be directed mainly at the wall with the amplitude of the direct signal kept to a minimum.

Figure 2 - Simple illustration depicting the concept of an image source, the grey handclap

Application of a Diffuse Field

A single conventional loudspeaker is a very coherent radiator and is relatively easy to localise. This is, of course, not desirable if we wish to disguise where the loudspeaker is. In order to achieve our goal of enveloping the listener while not surrounding them with speakers it is clear that the diffuse fields have some usefull attributes.

A diffuse field is not directional and is not localised by the listener. A diffuse field generated in front of the listener will consequently give a sensation of spacious envelopment.

Real environments produce ambient sound fields having diffuse characterstics largely determined by size, shape, and materials of the environment. Reproducing this ambient field involves mimicking their diffuse field. In a small room the presence of strong specular reflections makes this a difficult task. One approach is to use dipole loudpspeakers with electronic decorrelation achieved by frequency shifting [5].

The use of diffuse sources for surround channels is especially effective where the surround signal is mainly reverberant.

PRACTICAL SYSTEMS

In order to achieve the best result for a particular product type and price engineers must adopt a pragmatic approach using the most appropriate mix of techniques. We will consider four such realworld applications each one using some of the methods discussed.

A System Creating 'Virtual Loudspeakers'

One way of creating the surround effect is to use the HRTF derived filters described above to create the sound pressure at the listener's ears that would be produced by correctly positioned surround speakers.

The left and right (L&R) signals are fed to the appropriate speakers unaltered. The centre channel signal is fed equally to both left and right speakers producing a phantom image of the centre channel.

The surround signals are first modified to compensate for the crosstalk and then are filtered with the appropriate HTRF for a rear speaker position. Room reflections may also be simulated by adding delayed surround signals with some low pass filtering to give the effect of high frequency absorption by the walls. These modified surround signals are added to front signals on either side. [9]

A system Using Beam Steering to Create Wall Reflections A system using beam steering to create a surround sound from just one enclosure containing an array of drivers is disclosed in [10]. Beam steering requires the signal for each driveunit to be individually filtered. A number of beams are produced from the same loudspeaker array by summing the individual signals required to produce the beams.

To recreate the correct effect the center channel is radiated directly. The left and right are aimed so they impinge on the listener after one reflection whereas the surround impinges on the side-wall at a shallower angle being reflected a second time off the rear wall. Achieving this requires significant DSP processing power and 40 separate amplifiers & DA converters.

A System Using a Dipole to Create Wall Reflections

An innovative system in which a pair of loudspeakers each housing a pair of full-range cone driveunits positioned side by side using a specialised processor is disclosed in [11]. Each loudspeaker radiates as a monopole for the left centre and right signals. For the surround signals the processor inverts the phase of the signal to one drive unit producing dipole radiation.

The center channel signal is routed equally to both loudspeakers giving a central image. Where the loudspeakers are directed at the listener the listener will be positioned in the null where the direct signal is level low.In this position the listener will consequently hear mainly reflected sound.

A System Using Diffuse Dipole and Monopole Loudspeakers

A more recent approach disclosed by Dodd [12] uses a novel combination of diffuse dipole and coherent monopole. The diffuse dipole source is orientated with the null directed at the listener.

The diffuse dipole source may be most simply achieved by means of a Distributed Mode Loudspeaker (DML). The complex modal vibrations of a DML produce a strongly diffuse field since the radiation occurs from numerous regions of the panel with various phase and time delays. Opposite sides of the panel radiate with opposite volume velocity giving a dipole-like radiation pattern with nulls in the plane of the diaphragm [13].

The coherent monopole is provided by a coincident source loudspeaker directed at the listener in the usual manner. In this type of system the surround and front signals are applied to these different types of loudspeakers. No special signal processing is required. The left and right signals are applied to the appropriate coincident source loudspeaker. The centre channel signal is routed equally to both coincident source loudspeakers in the usual manner. Surround signals are suitably delayed by 10-20ms and then applied to the appropriate DML.

To achieve optimum envelopment by diffuse source the DML is orientated edge on to the listener. This results in the listener being positioned in the null of the diffuse dipole radiation produced by the DML [13]. This null is not as deep as the null in a conventional dipole, however the IACC is lowest in the plane of the DML [13]. It is believed that it is the low IACC at the listener position that allows the DML to produce a sensation of spaciousness without room reflections.

Some informal tests carried out by the authors have shown that a spacious sensation may be reproduced even in an anechoic chamber.

Another characteristic of the DML is that both the HF radiation and the IACC are greatest in the direction normal to the DML. This characteristic means that where a wall is suitably positioned specular reflections may produce virtual images to the side of the listener. In the ideal rectangular room these reflected sources may even surround the listener.

This approach results in unique combination of robust ambience, giving the sensation of being in the recorded environment, together with some reflected images where the domestic enviroment allows.

CONCLUSION

Home multichannel audio is no longer the territory of the enthusiast and as such consumer multichannel audio systems are evolving to meet the demands of the non-audiophile, non-enthusiast users.

Producing surround sound from fewer speakers inevitably requires some compromise in the sound quality and will necessarily be somewhat room dependent. Further compromises may also be found in cost and the region of the room in which the best sound may be experienced.

A number of different & ingenious approaches have been devised to solve the problem. Each approach has merits and areas of compromise. DSP, psychoacoustics and special transducer arrays all play a part in these solutions.

The resulting products extend the range of performance/cost/ergonomic compromises to the end user allowing more people than ever to enjoy surround sound in their homes.

REFERENCES

1. Wikipedia contributors, "Coherence (physics)," Wikipedia, The Free Encyclopedia, http://en.wikipedia.org/w/index.php?title= Coherence_(physics)&oldid=40928723 (accessed February 24, 2006).

2. J.R.Stewart, "The psycoacoustics of Multichannel Audio"AES UK "Audio for New Media" Conference

3. Hilmar Lehnert "Auditory Spatial Impression" AES 12th International Conference: The Perception of Reproduced Sound, 1993

4. D. de Vries, E.Hulsebos & J.Baan "Spatial Fluctuation of Spaciousness Measures in Auditoria" Presented at the 108th AES Convention 2000 Paris

5. Holman, T "New Factors in Sound for Cinema and Television," JAES, Volume 39, No.7/8, 1991.

6. Cashion.T, Williams.S, "Apparatus for creating 3D audio imaging over headphones using binaural synthesis" Patent US6195434 Filing date: 27.02.2001

7. Kirkeby, Ole; Nelson, Philip A.; Hamada, Hareo; "The 'Stereo Dipole': A Virtual Source Imaging System Using Two Closely Spaced Loudspeakers," JAES, Volume 46 Number 5 pp. 387-395; May 1998.

8. Nelson, Philip; Takeuchi, Takashi; "Optimal source distribution system for virtual acoustic imaging," presented at the 105th AES Convention, 4817, August 1998.

9. Chabanne, C "Method for improving spatial perception in virtual sound" World Patent WO 03/053099 A1, 26.06.2003

10. A Hooley, P.T.Troughton, A.G.Goudie, I.A.Bienek, P.R.Windle "Method and apperatus to direct sound"

11. J.R.Aylward "Percieved sound" US Patent 5,228,085, 13.07.1993

12. M.Dodd "Pistonically driven loudspeaker and perpendicular resonant panel loudspeaker in combined unit" UK Patent GB2392043 Filing date 18.08.2003

13. M N. Harris M.O.J. Hawksford "Measurement and Simulation Results Comparing the Binaural Acoustics of Various Direct Radiators" 107th AES convention, 1999 New York

www.kef.com

This article was first presented as a paper at the Audio Engineering Society's 21st UK Conference 'Audio at Home', April 2006.

www.aes-uk.org