VIC design session v22: reasoning

Design note, 18 July 2026

This sweep locked the cell geometry (18mm rod/34mm tube, 8-cell parallel, C_eff 1.37nF) and let the choke live-vary in turns/gauge while everything else held. All five top candidates cluster tightly at 303.89kHz — that's the resonant frequency set purely by L=100µH and C_eff=1.37nF, so screen-guiding turns/gauge with L pinned just reshuffles coil Q (3546 down to 1064) without moving f_r at all. That's the real lesson: I was sweeping the wrong lever. Turns count changes wire length and AC resistance, which drags tank Q from 27.0 down to 26.5 as coil losses rise, and cell voltage sags accordingly (324V → 318V) with series current following it down (847mA → 832mA). Higher coil Q buys me headroom but costs copper and doesn't touch frequency — the choke inductance itself is what needs to move, and 100µH is too small to get me under 100kHz with this capacitance.

Every candidate here trips the same violation: 303.9kHz sits outside the practical 1-100kHz band even though the vapor-film diagnostic is lighting up nicely inside the 80-250Td cold-chemistry window (129 Td at #1). I'm adopting #1 for now — best Q (27.0/3546.3), highest cell voltage (324V), least wire — since among these five it's strictly dominant, but I'm not satisfied with any of them.

Next session: sweep choke L directly (500µH-2mH range per my own notes) instead of turns/gauge at fixed L, to pull f_r down into band while preserving the CGDE-window hit.