MAR

THE PROBLEM

Recorded space rarely survives the system

A microphone array captures a real acoustic scene — an orchestra, a room, a crowd — in the relationships between its capsules. But those relationships were fixed by the array, not by the system that will replay them: a Decca tree, an ORTF-3D rig or an ambisonic mic each impose their own geometry, and mapping that onto a stereo pair, a 7.1.4 room or a WFS array is usually left to guesswork.

The usual answers are rigid. Fixed matrix decoders assume one target layout; ambisonic pipelines need high orders — and channel counts most systems cannot carry — to stay sharp, and they smear as soon as the capsules are spaced. The captured space arrives diffuse, the focus is gone, and the room the microphone was standing in leaks into every seat.

There is also the matter of scale. One capture may have to serve a stereo broadcast, a 5.1 archive, an immersive room and a WFS array — and re-miking for each is never an option. Whatever the array wrote down in a single take has to be enough to reach every layout that will ever play it.

HOW IT WORKS

How it works

UNDER THE HOOD

Going deeper

MAR treats the array as a description of a sound field rather than a set of channels. From the capsule signals and the known array geometry it reconstructs a compact spatial representation, then renders that representation on the target system — so the same recording plays correctly on two speakers or on sixty, keeping the width, depth and localization the microphone actually captured.

Three modes cover the range from a clean studio capture to a reverberant live room: a broadband mode for coincident arrays, an ensemble mode that resolves direction and diffuseness across the spectrum, and a focus mode for pulling a source out of a busy or reverberant scene. All run at a few samples of latency, with the per-band processing that spaced arrays need to stay coherent above their spacing frequency.

The focus mode is where MAR goes beyond decoding. Rather than a fixed beam, it measures how coherent the capsules actually are on the incoming sound and steers the image accordingly — the correlated, on-source energy is reinforced, the decorrelated room energy is held back. On a reverberant Decca or Hamasaki capture the source moves forward without the metallic artefacts a naive beamformer leaves behind.

The upshot is one capture, many destinations, with the spatial intent intact. A tree set up for stereo can feed a 3D room years later; a live array can be monitored in stereo at front-of-house and rendered in surround for broadcast at the same time, from the same signals — no re-patching, no second mix, and none of the depth the microphones were placed to catch thrown away.

AT A GLANCE

At a glance

PARAMETERS

Hands on the algorithm

POSITIONING

Compared to the alternatives

APPLICATIONS

Where it fits