Hypothesis: the floating-node ratchet (Scotchn Vbase, Chris CT common mode) is a diode clamp / charge pump fed through winding capacitance; capacitance couples edges but cannot bleed a DC level

Discord #neuralstan20-testing 2026-11-08. Scotchn notes that his circuit and Chris Freedom's are the same apart from the inter-wire capacitive coupling and the choke bypass. Both show a ratchet on a floating node: Scotchn's Vbase (49 V, falling to 0 at the start of a burst) and Chris's CT/plates, which step together by +15–20 V against CT (#3528). Scotchn had expected his ratchet to leak away through the mutual capacitance of the bifilar chokes. My reading: a capacitance passes only changes in voltage, so it cannot drain a DC level. Where both ends of the bifilar inter-wire C ratchet together, no charge moves through it at all. Capacitive coupling plus a one-way diode is the classic clamp / charge pump: each edge moves the node a fraction k = C_couple/(C_couple + C_node) of the remaining distance to the coupled peak, i.e. ΔV_n ≈ k·(V_peak − V_(n−1)). It stops climbing near the peak swing, less any resistive bleed (probe MΩ, leakage). The bifilar C therefore probably helps drive the ratchet rather than drain it. Chris's stall after about 4 pulses would fit k ≈ 0.4–0.5. That is an inference, not a measurement. Testable predictions: (1) the plateau level tracks the peak edge amplitude, not the DC drive; (2) adding a small known C (e.g. 100 pF) from the ratcheting node to reference makes the climb take more pulses but leaves the plateau about the same; (3) adding a bleed resistor lowers the plateau and makes it decay faster. Related: #3466 (open-loop stepping of stray capacitance draining through probe MΩ), #3528, #3536.