Flow-detection experimental record and bench tooling

Adds the design matrix and its supporting tools, and records in BRINGUP
what the 60-run matrix overturned: sub-40-percent duty mimics flow
(three of five dead-pump trials looked healthier than a working pump),
the operating point and threshold now rest on 25 pooled observations,
periodic re-checks are thermally free with the fans running, and the
settle gate closes a bench-proven miss. Also records what is NOT yet
validated - warm-loop baselines and behaviour under laser heating - as
first-light commissioning items.

scripts/bench/README.md

@@ -13,6 +13,10 @@ target board (dev image, python3 present) unless noted.
 | `fire_test.py` | FIRE drop-timing scope test (runs on the board): A = latch locked (expects nothing on FIRE/LASER_ON), B = latch unlocked / normal end-of-data, U = true underrun. Duty 0 throughout; refuses to unlock if HV reports good. All gates PASSED with it 2026-08-02 (2.0000 s pulses exact, both paths). |
 | `fan_test.py` | Fan/coolant bench (Windows-side): snapshots fan PWMs/tachs/temps, drives M8 → cut fans, M9 → cooldown → idle, verifying via tach readbacks. All-green 2026-08-02. |
 | `flow_characterize.py` | Coolant flow characterization using the factory temperature curve: baseline → flow → no-flow → recovery, printing the ΔT bands and their separation. Takes the heater duty as an argument (`flow_characterize.py 30`); aborts if downstream passes 45 °C. This is what showed the 10 % scheme was unusable (0.04 °C gap) and sized the 30 % check. |
+| `flow_matrix.py` | **The flow-detection design matrix** (with `flow_sampler.py`, which lives on the board at `/data/`): duty × duration × flow/no-flow, every run from a common cooled baseline, interleaved repeats. One heating trace yields the metric at every candidate duration, so cost and precision come from the same 60 runs. Prints a cost table, a precision table (mean±sd, worst-case margin, d′) and a ranked shortlist. Env: `FM_DUTIES`, `FM_REPEATS`, `FM_RESULTS`. |
+| `flow_sustained.py` | Long-run test of the real re-check cadence via M8: counts verdicts/false faults and tracks whether the loop accumulates heat. 40 min ⇒ 0 false faults, loop *cooled* 2 °C. |
+| `flow_warm_validate.py` | Runs the real check from a heater-warmed baseline. Note the ceiling: 100 % duty pushes the downstream sensor past 50 °C in 30 s while the bulk barely moves, so warm-loop validation above ~23 °C needs the laser, not the heater. |
+| `flow_recheck_char.py` | Characterizes short in-run re-checks and the differential metric; documents two dead ends (over-temp cannot see a stopped pump; passive warming trends are ambiguous). |
 | `temp_calibrate.py` | Coolant temperature calibration helper (`watch` / `point <measured_C>` / `fit`) — pairs a measured temperature with averaged raw ADC readings and fits the line. Used to sanity-check the factory curve against a thermometer. |
 | `build-glowforge.sh` | Cross-compiles **grblHAL-glowforge** (the canonical driver repo, `../../../grblHAL-glowforge`) in the forge-yocto WSL distro. Run: `wsl -d forge-yocto -- bash <path>/build-glowforge.sh` (from PowerShell; Git Bash mangles /mnt/c paths). This is the production controller build. |
 | `build-feeder.sh` | Cross-compiles `feeder.c` the same way. |