Skip to content

Acceptance Tests

Conditions (all tests)

Machine V3 prototype, calibration of 2026-11-17, settings as committed on 2026-11-18 (config.toml, printer.cfg)
Board Test PCB (Software/Slicer/tests/fixtures/Test PCB/), top side, held in the PCB mounts
Job jobs/Test PCB-top-0603.job.json: 23 x 0603 resistors
Parts 0603 resistors on 8 mm paper tape in feeder F1
Tip N08 (board touch Z 39.5, inspect Z 34.6)
Air compressor ~60 psi
Room indoor lab, normal ceiling light; machine lights as configured (frame strip 15 %)
Operator steps allowed load the board, load tape, Vision home, Start; answer dialogs

Requirements and tests

# Requirement Test Pass criterion
1 Operate without a laptop or network: start, stop, jog, home, choose a program on an HMI load a job from an SD card / USB stick and run it using only the touchscreen job loaded, started, watched, stopped from the screen
2 Vision identifies passes and fails correctly every up-camera decision (PASS / REJECT / EMPTY) and every board-check decision (PLACED / MISSING) over the runs, judged by a person from the saved photo (human_agrees column) ≥ 95 % correct out of ≥ 100
3 Pick 0603/0805 parts from tape picks that came up with a good part / all picks, from the event log (report) ≥ 95 %
4 Feeder runs without failures consecutive 23-part runs from F1 runs complete; each miss logged and retried
5 500 components per hour seconds per part over a full job, timed by run-job (alignment excluded) ≥ 500 CPH (≤ 7.2 s/part)
6 Placement accuracy (characterised) check-board after each run: offset of each part from its pads reported (mean, worst)
7 Continuous operation four 23-part runs back to back without repair, re-calibration or reboot all complete

Procedure

  1. Daily startup. Load tape in F1. Empty the reject bin.
  2. Put a bare Test PCB on the plate.
  3. Touchscreen -> Job -> Test PCB-top-0603, Align first on -> Start.
  4. When done, run python -m pnpvision check-board "jobs/Test PCB-top-0603.job.json".
  5. Replace the board, repeat for the next run.
  6. After the last run: python -m pnpvision report --out report_latest.
  7. Fill in human_agrees in vision_review.csv and placement_review.csv by looking at each photo; count the agreements.
  8. Back up pnp_events.sqlite, report_latest/ and results/.

Results: Results.