4.1 Functional Testing
Power, signals, and performance checks.
Functional testing is where a silent build turns into a proven product. It applies power safely, drives I/O with realistic loads, and compares real measurements to defined limits—no guesses, just numbers. Recipes and fixtures make the steps repeatable, while a calm bring-up protects hardware from inrush and wiring mistakes. Results are tied to the serial number (often through MES—manufacturing execution system) so performance, firmware, and proofs travel with the unit. Do this well and test becomes a quiet gate: predictable on the line, and a strong shield against field surprises.
4.1.1 The mission (in one line)
Turn a silent box into a measured, proven product—apply power safely, exercise the I/O with realistic loads, compare against limits, and record the truth to the serial number.
4.1.2 Test flow that keeps you safe (and fast)
Plan → Power → Exercise → Measure → Decide → Record
- Plan — load the right recipe by SKU/Variant scan; fixtures/loads ready.
- Power — safe bring-up (current-limit, sequencing) with interlocks.
- Exercise — drive interfaces and sensors like the real world.
- Measure — voltage, current, timing, comms, thermals vs limits.
- Decide — PASS/FAIL by rules (no “it looks fine”).
- Record — push results, plots, and versions to MES by SN.
4.1.3 Fixture & station essentials
- Interlocked lid, E-stop, bleed/discharge for caps; HV/HOT beacons.
- Mating connectors (prefer over pogo) sized for current; strain-relief for harnesses.
- Programmable PSU & loads (constant current/voltage/resistance; fast foldback).
- Sensors/actuators simulators: thermistors, 4–20 mA loops, relays/solenoids, encoders.
- Network hub with known-good links; RF attenuators or shield box if radios are checked.
- Calibration assets: DMM/clamps with current certs; station self-check at shift start.
- Recipe control: test script and limits selected by scan; block manual edits.
4.1.4 Power-up sequence (don’t cook first)
- Visual pre-check: fans free, jumpers/config switches per SWI, no loose screws/FOD.
- Set limits: PSU current limit slightly above spec; electronic loads disabled.
- Apply rails in order (if required): e.g., 5 V → 12 V → 24 V; verify inrush < limit.
- Brownout test (quick): dip primary briefly; product should not latch in a bad state.
- Thermal sniff: 2–5 minutes at idle; fans spin; no hotspot beyond design expectations.
Abort if over-current trips, smoke/odor, abnormal noise. Quarantine with a ticket.
4.1.5 Signals & interfaces (exercise like the real use)
4.1.6 Performance checks (numbers that matter)
- Voltage drops under load at key test points (spec’d rails).
- Current draw: idle / typical / max; compare to golden.
- Timing: boot time, relay actuation delay, watchdog behavior.
- Throughput: network bandwidth, serial packet loss, USB speed.
- Thermals: ΔT at heatsinks after a brief load; fans meet RPM window.
- Noise (optional pre-compliance): ripple on rails, simple radiated sniff in fixture.
4.1.7 Limits & golden references
- Limits live in the recipe, versioned and locked.
- Build golden unit data sets (current, drop, timings) at NPI and after ECOs; use tolerance bands (e.g., ±10%) unless the spec dictates absolute numbers.
- Separate spec limits (must meet) from golden trends (drift detection).
4.1.8 Pass/fail logic (no vibes—just rules)
- Hard gates: safety, over-current, miswired I/O, failed storage, watchdog faults → FAIL.
- Measured gates: numeric compare to limit table → PASS only if all required rows pass.
- Retest rules: one immediate retest allowed only after a clear cause (fixture reseat, cable change). Second fail = NG-QUAR.
4.1.9 Environmental sanity (fair tests)
- Log ambient Temp/RH; some limits vary with environment.
- If the spec requires warm-up or soak, the script enforces wait times.
- For radio checks, use a shielded mini-box or attenuators to avoid the open lab RF zoo.
4.1.10 Data & traceability (what the record must contain)
Attach to the unit SN:
- Recipe ID & version, fixture ID, instrument IDs (cal status).
- Limits table used; operator and timestamp.
- For each test: stimulus (V/I/timing), measured values, PASS/FAIL, and plots where helpful (e.g., current vs time).
- Firmware image/hash, config/keys written, and results of any checksum verify.
- Photos if your SOP wants panel/label proofs at test.
4.1.11 Throughput & ergonomics (takt-aware testing)
- Pre-stage two fixtures (ping-pong) if test time ≈ takt×2.
- Parallelize long soaks off the main lane (side rack) and return for closeout.
- Use quick-connects and guided nests; no raw pin probing.
- Make common failures obvious on the UI (red tile with the failing step, not a log file hunt).
4.1.12 Starter test matrix (customize per product)
4.1.13 Acceptance cues (fast eyes)
4.1.14 Common traps → smallest reliable fix
4.1.15 Pocket checklists
Before
- SKU/Variant scanned → correct recipe/limits loaded
- Fixture interlock/E-stop OK; instruments in cal
- Loads zeroed; PSU current limit set; fans clear
During
- Power-up clean; inrush OK; no alarms
- Exercise I/O per script; measure and log automatically
- Quick thermal sniff at load; fans meet RPM window
After
- PASS/FAIL decided by rules; no manual override
- Results, plots, firmware hash bound to SN
- NG units to QUAR/REWORK with failing step noted