Can a backbeat-heavy mix pull recipe v5 off the downbeat? No: the catch map, the kick-removed remixes and the real phase errors all hold the fence

Listening · the failure study

Study Overview

The Ask and the Fence Arithmetic

The app's brief did its own analysis before asking: with the slide fenced at half a beat, a phase error under half a beat can only reach the downbeat, whatever the backbeat weighs, and a phase error beyond it can only land a beat late. The study's job was to confirm the geometry, find how big real phase errors get, and probe the window's interior. The geometry confirmed exactly: the catch map holds the downbeat for errors within about 0.45 beat at every weight tried, the 0.05 shaved at the edges being the sweep grid and the 10 millisecond comb step, and nothing ever landed anywhere but a beat.

Real Material Never Gets Close

The gated arm rebuilt each song's mix with the drum stem high-passed at 150 Hz, a fourth-order skirt that removes the kick and keeps the snare and hats, the backbeat-heavy construction the ask feared. All four songs kept their Study 27 standards: letter-lands ends 0.04 to 0.06 seconds with a 0.051 middle, peaceful-dreams 0.02 to 0.07 with 0.042, mop-heart exact, eat-all-the-clocks at its usual quarter-bar edge on the corrected chart, and a control confirmed the full-mix runs reproduce Study 27's deltas within 2 milliseconds. The measured phase errors the slide actually corrected are 0.05 to 0.10 beats. The ask's 1.4-beat arithmetic for mop-heart was right and its premise was not: that 0.38 seconds was the superseded truth anchor's error, the ruler, not the alignment's phase, which sat 0.05 beats from the lattice. One unit correction went back in the answer: an eighth note is half a beat, so the proposed pickup sits at the fence boundary, not inside it.

The Knife-Edge, and Whose Remedy It Is

The in-window arm produced the only captures: against a downbeat at 0.3 weight, a 0.8-energy pickup or matching hi-hat off-eighths take the slide to the fence edge, half a beat off. These are boundary ties on synthetic impulses, never approached by the kick-removed real mixes, so the registered conditional remedy arm did not run and v5 ships unmodified. The remedy the ask proposed, scoring the grid against itself shifted by whole beats, is still the right instrument for the construction that DID get captured the same morning: Study 28's chartless anchor, which has no DTW tether and landed on the backbeat on two songs. The failure study and the capstone study answered each other within hours, and the downbeat check moves to the auto-chart feature's queue.