Experimental Audio Systems — Custom Synthesis Hardware & Wave-Terrain Instruments

This division builds instruments that do not yet have a product category. A single prototype may contain a discrete transistor VCO, a Teensy 4.1 Cortex-M7 supervisor, a Dante/AES67 network bridge and a machined aluminum acoustic horn — all tuned together, not partitioned into “hardware” and “software.”

Wave-terrain synthesis, physical-modeled resonators and multi-head tape transports recur across the archive. PROT-007’s kinetic percussion automaton and PROT-011’s spinning speaker array translate abstract DSP into moving air with millimetre tolerances. Every bill of materials lists supplier, part number and tolerance because reproducibility is part of the research claim; every acoustic profile is measured on-bench before being offered as a playable preset.

Work is staged between Studio B (Physical Computing) and the field vault, where 64-channel Genelec arrays and laser-Doppler vibrometry validate dispersion and inter-channel phase coherence. Patents in this cluster are defensive publications documenting topologies — e.g., multi-vector joystick morphing of neural latent spaces — so proprietary audio conglomerates cannot enclose them.

Synthesis topology is documented at the same fidelity as analog circuitry. Wave-terrain maps are published as 512×512 normalized float arrays with explicit trajectory equations and boundary conditions, so a reader can re-render the instrument without our hardware. PROT-007’s kinetic percussion and PROT-042’s neural latent resynthesizer each include a WASM audio profile with carrier, modulation, filter and binaural delta calibrated from bench — not a preset approximation. This is why the Acoustic Bench can credibly audition a prototype before it is built.

Calibration is field-anchored. After bench validation, 32- and 64-channel diffusion maps are transported to the Vault for inter-channel coherence tests; phase cancellation greater than 3 dB at any seat triggers a redesign of the fractional delay staging. Results, including failed staging matrices, are logged as research notes and, when topologies are novel, as defensive patents (PAT-2021-007 series). The division’s publication pipeline is therefore not “demo → paper” but “bench → vault → field → disclosure,” each with measured tolerances.