VIC design session v93: reasoning

This sweep did not do what I set out to do, and that's worth recording plainly. I asked for a 1.37nF cell (1-2 tubes) driven up into the untested 32-50kHz band. What the scorer actually surfaced is the opposite corner: 8 tubes in parallel, C_eff = 20.53nF, resonating at only 12.91kHz. The 3.7mH chokes I specified pair with that fat capacitance to drop f_r well below my target band, so the "high-turn coils into 32-50kHz" premise never materialized. The optimizer chased CGDE window compliance instead, and every top candidate parks the vapor-film at 140-186 Td — squarely inside the 3-6 eV cold-chemistry window. That's the real signal here: score is being driven by film physics, not by frequency or Q.

And Q is the disappointment. Tank Q = 1.2 across the board — the 509Ω cell series R plus 1203Ω bulk water swamp the 600Ω reactances, so there's almost no voltage magnification (×1.2). Coil Q of 133 is wasted; the water, not the copper, is the loss. That's the central trade-off: I can push Td by raising drive voltage (candidate #4/#5, 467V → 186 Td) but breakdown margin falls ×85→×64 and circulating current climbs to 0.78A.

I'm adopting #3: 350V drive, 5% duty, 29mA average input, 140 Td, margin ×85. Same film chemistry as #1/#2 at a fraction of the average current — duty is the free lever.

Next session: shrink to 1-2 cells to actually cut C, lengthen chokes to 7-9mH, and force f_r into 32-50kHz so I can finally test whether Q survives at higher frequency.