Fenophone measured spectral landscape: sweep, nulls, and microtiming datasets + figures

The measured spectral landscape of the Fenophone instrument: the committed control-sweep, surrogate-null, spectrum, ensemble and microtiming datasets, plus every figure rendered from them (PDF and PNG, in the accompanying archive).

What is here. sweep-v1/v2/v3 and their gate-point companions: the DFA α and multifractal-width response of the emitted onset stream over the instrument’s primary control plane, at pinned seed panels. nulls-v1/v2 and the nulls-e2-spread500-* family: each pack’s scaling statistics scored against its own IAAFT and shuffle surrogates. spectrum-v1: the full Legendre spectrum on onset streams over q ∈ {−4…4}. ensemble-v1: per-voice, aggregate and consonance-rank series for the multi-voice path. microtiming-dfa-v1: the humanization benchmark against measured human drummers with a phase-randomized regime null. Each row pins the scaffold_version it measured.

Selection caveat on the catalogue-wide IAAFT result (nulls-v2)

The nulls-v2 dataset records all eleven shipped packs beyond their IAAFT surrogate null, where the earlier four-pack nulls-v1 recorded three of four. That improvement is not a blind survey: the measured retune campaign that closed the soft-genre misses selected packs on the IAAFT-excess statistic itself, so the catalogue-wide 11/11 reading is a post-tuning result and must be quoted as one. Each dataset pins the scaffold_version it measured, so every row stays exactly true of the versions it names.

Two further scope notes travel with these datasets. A null width-vs-IAAFT result is a no-verdict about nonlinearity, never evidence of linearity. And the word “width” denotes two different statistics in this program: the two-point MFDFA width h(−2) − h(2) of event-time onset streams (the claim-bearing statistic these datasets score, typically 0.08–0.29) and the Legendre spectrum width Δα of audio amplitude envelopes (typically ~1.3). They differ by roughly 10× in range and a number from one is meaningless against the other’s nulls or bands.

Part of the Feno program. The source repository is private; the reproduction materials for this record are contained in the record itself.