analog M5f: the tracking switch hold step and droop from the PSP103 tables with charge and leakage (first lane run as a T27 OS pane) #11378

Part of the analog epic (see #10486); the first lane run as a T27 OS pane on the free key tier (owner's routing, #11317; permission mode A, #11362). After #11374 the acceptance evaluator gives 8 PASS, 0 FAIL, 5 NOT MODELLED of 13 blocks. The tracking switch block is NOT MODELLED for two rows because its model (specs/analog/sar_switch.t27) is square law: no gate capacitance, so no charge injection (hold step), and no leakage (droop).

The references are already in the repository: our own ngspice-47 + PSP103 run of the same cell (specs/analog/decks/accept_switch.cir) gives a hold step of -1.596 mV and -1.730 mV at the report's two conditions (the report: -1.6 and -1.7 mV; limit +-30 mV) and a droop under 10 uV over 413 ns (limit +-5 mV). The tt PSP103 tables now carry drain, gate and source charge (sar_comparator_q.t27 sq_* functions, mos_psp103_inv20.t27) and the log-domain current read for weak inversion (mos_psp103.t27).

Requirements

• FR-001: the hold step at both conditions MUST be computed from the tt tables with the devices' charge (the switch's gate edge injecting channel charge onto the hold capacitor), and compared with -1.596 / -1.730 mV, the gap printed; the row is PASS only inside the report's limit AND within 10 % of ngspice.
• FR-002: the droop over 413 ns MUST be computed from the off switch's leakage (the log-domain current read) and compared with ngspice's "under 10 uV"; PASS only inside the limit and consistent with ngspice's bound.
• FR-003: accept_sar.t27 and docs/analog/accept-sar-2026-10-20.md recompute the tracking_switch block and the totals.
• FR-004: negative controls: the charge removed gives no hold step and is refused; a leakage ten times too large is refused.

Boundary

• specs/analog/ (a new spec for the switch with charge if needed; existing specs only by extension)
• specs/analog/accept_sar.t27
• docs/analog/accept-sar-2026-10-20.md

Success Criteria

• t27c test-report on every changed spec prints FAIL 0 and vacuous passes 0
• the hold-step and droop rows are PASS, or the measured gaps are written and the rows stay NOT MODELLED with the reason

🤖 Generated with Claude Code