Scotchn single-diode bifilar test: diode removed rings 541.53 Vpp vs 86.7 Vpp with diode in (same 4 V drive); CH1 probe-ratio error voids earlier '0.9 V forward drop' reading

Bench: Scotchn, Discord #neuralstan20-testing 2026-09-29. Setup: UU nanocrystalline core (clamped, 1 layer Kapton shim), primary 10T 0.249 mH (100 kHz LCR). Bifilar pair A/B, each 3.8 ohm / 33.3 mH at 100 kHz; A-B interwire capacitance 1.18 nF. Diode sits between the end of A and the start of B. Test load is two 2A102J (1 nF, 100 V DC, PET) in parallel (~2 nF) across the outer ends. Drive: 20 kHz, 50 %, 9-pulse bursts, gate 1500, supply 4 V. Probes: CH1 and CH2 are 100x differential, CH3 is 1x passive (MOSFET signal).

• Results:
• (1) Diode in — cap shows double humps, 86.7 Vpp, Vmin -5.55 V; the first hump is about 22x supply, which matches the turns ratio. CH1 read 879 mVpp with its probe ratio left at 1x, so the real diode voltage was about 88 V. My earlier reading of this as a ~0.9 V forward drop, and my 'hump 2 is conduction' conclusion, are VOID.
• (2) Diode removed (open), same 4 V settings — cap rings bipolar at 541.53 Vpp (Vmin -205.36 V), growing through the burst, then free ring-down at about 8.8 kHz (read off the screen at 200 us/div) lasting more than 700 us. So the diode position is not open to AC; a large capacitive path bridges it. The diode in circuit cuts cap swing about 6x.
• The 8.8 kHz does not fit my simple lumped models of the interwire-C path (those give roughly 15-32 kHz); mechanism unresolved. The 100 V caps were overstressed about 5x during run (2).
• Open — re-run diode-in with CH1 at 100x and known polarity; diode-shorted run; lower-supply diode-out run.