WFC design session v21: reasoning

Lab notebook, 4 July.

Fine-resolved the geometry around candidate #1 with gap/length variation at fixed chemistry (0.1ppm Na+, 3°C, ×0.25 surface treatment). Five candidates cluster tightly, 0.832–0.840, and the trade-off is subtle rather than dramatic: widening the annular gap (e.g. #5, 8.5mm/12.0mm) drops C_eff, which pulls f_r down toward 95.7kHz and gives a bit more core-loss headroom in the choke, but it costs tank Q (97.6 vs 100.7) and breakdown margin (×100 vs ×111). Narrowing the gap does the opposite. Series current is essentially pinned at ~982-984mA across all five — we're riding right at the <1A ceiling regardless of geometry, so that constraint is chemistry-limited (Randles series R ~11.1Ω), not geometry-limited. λ_D and C_dl barely move either, confirming the dielectric double-layer is stable across this gap range.

I'm adopting #1: 9mm rod / 13mm tube / 100mm, single cell. It sits closest to my 100kHz target (98.84kHz), gives the best Q of the set (100.7), and 1208V cell voltage from 12V drive is the strongest gain — the ×111 breakdown margin is still comfortable even though it's the lowest of the five. The thermal margin question wasn't fully resolved by geometry alone here; gap widening buys core-loss headroom but not enough to justify the Q/voltage sacrifice at this purity level.

Next session: sweep choke turns/wire gauge directly for thermal margin at fixed #1 geometry, and test higher-ppm water to see if series R (not geometry) is the real current bottleneck.