Ambiophonic Principles for the Recording and Reproduction of Surround Sound for Music Angelo Farina (1), Ralph Glasgal (2), Enrico Armelloni (1), Anders Torger (1) (1) Industrial Eng. Dept., Università di Parma, Via delle Scienze – 43100 Parma - ITALY E-MAIL: farina@pcfarina.eng.unipr.it - HTTP://pcfarina.eng.unipr.it (2) Ambiophonics Institute, 4 Piermont Road, Rockleigh, New Jersey 07647, USA E-MAIL: glasgal@ambiophonics.org - HTTP://www.ambiophonics.org This paper discusses the psychoacoustical background and the computational issues involved in the real-time implementation of a complete Ambiophonics reproduction system based on binaural technology. Ambiophonics, which requires only two media channels, evolved from previously known approaches such as the reproduction of binaural recordings over closely spaced loudspeakers through cross-talk cancellation, and the reconstruction of hall ambience by convolution from suitable impulse responses. The equations for the design of the digital filter coefficients are derived with regard to the many possible kinds of pre-existing recordings (binaural, sphere, ORTF, M/S), and their implementation on available hardware and software platforms are described. The authors suggest psychoacoustic explanations for the perceived audible performance, and describe the first results of a comparative listening test, evaluating the realism of three periphonic surround reproduction systems: Stereo Dipole, Ambisonics and Ambiophonics. 1. INTRODUCTION In recent years many different surround reproduction systems have been developed. Many of them, such as 5.1, are spatially limited and are not considered to be psychoacoustically valid methods for achieving a realistic reproduction of recorded music. Other paradigms that are capable, in principle, of complete periphony (reproduction of apparent acoustic sources everywhere in the space, over a complete sphere around the listener) have been proposed, but none of them has gained acceptance or become commercially available on the market, in part, because they were not compatible with the vast existing library of two channel LPs and CDs. In this paper only potentially complete periphonic systems are considered, with the goal of providing the mathematical description and the implementation details of one of these methods, termed Ambiophonics; it will be shown that this method, used only in reproduction or used in both recording and reproduction, makes use of physical principles and digital filtering techniques, which are also found separately in other periphonic surround methods, but are here coupled together in a consistent and psychoacoustically correct form. Most periphonic methods fall in one of two broad categories, summarized here: A) Binaural methods: the sound field is originally recorded with some sort of dummy head microphone, and reproduced by delivering the recorded signals unaltered to the entrances of the ear canals of the listener. B) Wavefield reconstruction methods: the system replicates the wavefronts, impinging on an array of microphones in the original space, through the use of coarse or dense arrays of loudspeakers during reproduction in a different space. The binaural methods have the advantage of requiring the recording and transmission of just two channels, and are thus compatible with traditional two-channel stereophony, a form of monodimensional reproduction (virtual sound sources located on a line). Possible methods of replicating the recorded acoustic pressure signals at the ears of the listener include headphone reproduction (with or without head tracking) and loudspeaker reproduction. In the latter case, cross- talk cancellation is usually required, for canceling the spurious signals that go to the “wrong” ear. The common binaural methods are very sensitive to both the shape and directional characteristics of the original microphone employed for the recording and of the particular human head of the listener [1]. If these two characteristics do not match well, the spatial illusion is poor and the reproduced sound field is judged unnatural. Furthermore, localization errors are common, particularly for sound sources near the median plane (front-back confusion, height