TouchProof

The inverting mechanism is dynamics-from-kinematics: force equals effective mass times deceleration, so the pitch reads deceleration from the CSI motion envelope and multiplies by a mass model. Two independent kills. First, path-length modulation carries no force information — a 30 N brush and a 150 N clamp can produce identical motion envelopes because effective mass, contact stiffness, and pose are unknowns RF cannot observe, so the inversion is underdetermined at any sampling rate; this is the band's rule 3 stated as physics rather than policy. Second, the contact transient occupies milliseconds and structural bands far above the 3.3 Hz ambient CSI Nyquist (0.41 Hz on the decimated WASM path), so the deceleration the model needs is never sampled — this kill holds even granting a perfect mass model, which makes the two kills independent. A third bar is structural rather than physical: verifying force limits is a performance claim against the certified stack's own protective function, barred by R-A1 regardless of what the physics could do, and injury-severity inference is barred with it. The salvageable sliver: the PDM mic honestly detects an impact as a broadband level transient — 'a contact event occurred at T' can enter the session ledger and the cell certificate as an event marker with no severity attached, and that sliver already ships inside the cycle-conformance and phantom-actuation rows, where its claim stays within one sensor's physics.