VIC design session v58: reasoning

This sweep didn't do what I asked, and that's the finding. I set out to push C toward 5–7nF with 1–4mH chokes to seat f_r in the 40–70kHz band. Instead the optimizer walked every top candidate back to the 100µH anchor with an 8-cell, 1.37nF stack, parking f_r at 303.9kHz — a flat violation of the practical band on all five. It did this because the scoring rewards the CGDE window, and the tiny chokes keep coil Q enormous (3546 down to 177) while the film sits comfortably at 113–129 Td, dead center of the 3–6 eV cold-chemistry band. So the score climbed while the buildability collapsed.

The real trade-off is stark: coil Q is bought almost entirely with wire gauge and turns at fixed 100µH, and every candidate clears the 700mA floor (847 → 740mA), so current restriction is NOT the binding constraint here — frequency is. The 100µH anchor plus 1.37nF simply cannot resonate below 100kHz; √(LC) forbids it. The optimizer found a local maximum by cheating my frequency rule.

I'm adopting #1 only as a diagnostic reference, not for build — its coil Q of 3546 and 96.5% displacement current confirm the cell itself is nearly ideal. But I will NOT wind hardware to a 304kHz design.

Next session: lock the 1–4mH chokes as a hard floor (drop the 100µH anchor entirely), and force C up to 5–7nF by going to 16–24 cells or larger tubes, so LC actually lands 40–70kHz. Then re-check whether the film still lights in-window at the lower per-molecule voltage that larger C implies.