Header image  
by Ralph Glasgal
line decor
Home Tutorials Tech
Kudos and
Demos Bio Free Ambio
Glossary The Home
Concert Hall
Rec Engineers
FAQ/Forum Links Contact us
line decor
e WIFR Structure

Ambiophonic Principles for the Recording and Reproduction of Surround Sound for Music -Part 2

Angelo Farina, Ralph Glasgal, Enrico Armelloni, Anders Torger


The goal of the Ambiophonics reproduction method is to create a realistic listening experience starting from existing 2-channel or even 5.1 recordings. Fortunately, the recordings themselves are not usually predistorted by the stereo reproduction process. That is, the recordings do not contain crosstalk and do not know that they will be listened to via a stereo speaker triangle that engenders crosstalk, requires phantom imaging rather than binaural localization, generates comb filtering and introduces pinna/HRTF angle errors. Normal recordings typically include very limited "3D surround" information. Of course, the missing information must be recreated in some way: this is done by means of convolution with suitable room impulse responses.

The method can be basically explained as the superposition of two simultaneous periphonic reproduction systems: cross-talk cancelled reproduction over a pair of closely-spaced loudspeakers (as is usually done for binaural loudspeaker reproduction), and approximate wavefront reconstruction with an Ambisonics array, being fed with reconstructed hall ambience signals derived from the left and right direct sound disc channels convolved with a set of weakly-correlated real hall impulse responses.

Figures 1 and 2 show the basic scheme of the two parts of the system.

Fig. 1 Stereo-dipole reproduction through
cross-talk canceling digital filters

The crosstalk cancellation operation is performed through the convolution of the two input signals with a set of 4 inverse filters, computed taking into account the kind of microphone employed for the recording. These inverse filters "cancel out" a great part of the microphone-dependent spatial effects (for example, the particular pinna coloration of a dummy head, in the case of binaural recordings), and thus leave each listener capable of hearing "with his own ears". Of course, this deconvolution is easiest if the microphone did not introduce very sharp filtering curves: in fact the Ambiophonics reproduction is usually better starting from recordings made with a "pinna-less" dummy head (sphere microphone, or ORTF microphone). In principle, any kind of stereo microphone can be used, even a "virtual" one, as happens when the stereo mix is obtained by level panning many monophonic recordings of various instruments or vocalists. Thus there is almost no two-channel recording that does not benefit from being reproduced Ambiophonically.

In the following section, the mathematical details for the derivation of the cross-talk canceling inverse filters will be described.

Fig. 2 Virtual Ambisonics reproduction by convolution with two sets of 3D impulse responses.

The surround loudspeaker array is responsible only for the reproduction of off- stage early reflections and reverberation tails. This means that the direct sound must be deleted from the impulse responses employed for convolution. In principle, these impulse responses can be obtained from Ambisonics decoding of a single B-format impulse response, synthesizing many virtual coincident (hyper)-cardioid microphones, each of them pointing towards a loudspeaker. But in practice it is preferable to consider these impulse response as an undersampled set of Wave Field Synthesis impulse responses, obtained by non coincident microphones, placed at relative positions from the main stereo microphone corresponding to the relative position of the reproduction loudspeaker from the listener. Following the WFS theory, the directivity pattern of each of these displaced microphones should depend on the directivity pattern of the loudspeaker being fed by its signal, but in practice this factor has been found to be very subtle, and can be neglected in most cases (provided that the loudspeakers employed for reproduction do not exhibit very strange directivity patterns or are in the presence of close reflecting surfaces).

In section 4 it will be shown what numerical processing is required to adapt the three-dimensional IRs, measured in a concert hall, to make them suitable for use as filters in Ambiophonics processing.

Section 5 presents available implementations of Ambiophonics, based on currently available hardware platforms, and on forthcoming softwareonly solutions, based on advanced convolution algorithms which have been recently implemented for real-time operation on low cost PCs.

Finally, subjective comparative tests were performed, in which it was possible to assess the preference for one of three simultaneous recording/reproduction methods: Binaural, Ambisonics and Ambiophonics. The tests were performed in a special listening room, equipped with a configurable reproduction system. At the time the tests were conducted, it was not possible to perform real-time reproduction of random recordings, but for comparative tests it was easy to pre-compute all the required signals, and leave the listener free to switch among the three systems.

The results of the subjective tests indicate that the Ambiophonics system is significantly preferred over the other two, followed by the binaural method (Stereo Dipole) and by Ambisonics. It was also confirmed that the "synthetic" Ambiophonics reproduction (in which the surround channels are derived by convolution from the main two channels) is almost indistinguishable from the "true" (directly recorded) surround obtained by processing the Bformat recording. This is what makes it possible to obtain such satisfying results from existing twochannel recordings, although of course when 4 or more channels are available, it could be preferable, depending on the source material, to reproduce the frontal pair over the Stereo Dipole and employ the rear pair for convolution with the surround impulse responses.

<<Previous Page | Next Page >>

Article Pages 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | References