WFC design session v32: reasoning

This sweep was supposed to land the vapor-film inside the 80-250 Td cold-chemistry window by dropping cell voltage. It didn't. Every top candidate lights the 0.1mm film at ~318-319 Td in the townsend-avalanche regime — above the CGDE band, not in it. The scoring saturated at 0.624 because the resonant/Randles floor is clean (breakdown margin ×244, displacement current 97.9%, no violations) but the film diagnostic is not rewarding us. That's the honest headline: I hunted the window and missed it high.

The trade-off is stark. To resonate near my 45kHz target I need C_eff ~2.65nF, which forces 18mm-in-24mm or 14mm-in-20mm geometry — and that same geometry, at Q=22.3 off a 36V drive, magnifies to 801V across a 3mm gap. 801V over a 0.1mm film is ~319 Td no matter how I slice the duty. Frequency actually settled at 97.7kHz, not 45kHz — the small C pulled f_r up. So Q, band, and Td are coupled: I can't lower Td without either lowering Q (less magnification) or widening the film assumption.

I'm adopting #1 (18/24mm, 2% duty, 26mA average input) — lowest average current for identical score, best leakage economy while I'm still outside the yield window anyway.

Next session: stop chasing C. Drop Q to ~10-12 and drive harder, or model a thinner 0.05mm film, to pull cell voltage toward ~250-400V and walk Td down into 80-250. Also sweep gap below 3mm explicitly — that's the untested lever on the Td denominator.