Scotchn single-diode bifilar test, properly probed: diode blocks ~87 V during BOTH humps; cap humps are displacement-coupled; diode clamps the opposite overshoot (amends #3445)

Bench: Scotchn, Discord #neuralstan20-testing 2026-09-29/30. Same setup as #3445:

• UU nanocrystalline core; primary 10T.
• Bifilar A/B, 33.3 mH each; 1.18 nF A-B.
• Diode between A end and B start.
• ~2 nF PET test cap across the outer ends.
• Drive: 20 kHz 50 %, 9-pulse bursts, 4.00 V supply (0.000 A shown).
• CH1 = 100x differential across the diode (verified connected this time; an earlier CH1 trace was an unconnected lead picking up signal like an antenna). CH2 = 100x differential across the cap. Both at 50 V/div.

RESULT:

• CH1 (diode) and CH2 (cap) show the same double negative-going humps, about 87 V deep, time-coincident, flat about 0 V between pulses and after the burst. No free ring after the burst.
• A conducting diode would read about 1 V, so the diode is NOT forward-conducting during either hump; it blocks about 87 V.
• The cap's humps therefore arrive by displacement current through the capacitance bridging the diode (interwire C plus the diode's own capacitance). This supports Scotchn's claim: conduction blocked, capacitive path remains, and the second same-polarity hump happens with no diode conduction.
• Compared with diode-out (#3445: 541.53 Vpp bipolar ring at about 8.8 kHz), the diode kills the opposite-polarity swing and the free ring. That suggests it conducts only to clamp the reverse overshoot ('negative overshoot rectified', per Scotchn).
• Caveats: probe polarity (which diode lead is +) not yet stated. Hump 2 still peaks inside the on-time, so it is not shown to be the pulse-end collapse event of US 4,936,961 / 5,149,407. The equal split between diode and cap suggests an effective bridging C about equal to the ~2 nF test cap (a capacitive-divider hypothesis, untested).
• Proposed tests: flip the diode (it should then conduct on the humps, and the cap should hold charge); use a single 1 nF test cap (the divider model predicts cap share rises to about 2/3).