owui-tools/docs/testing/prompt-corpus/hard/h03-pcb-design-review/rubric.md

h03 — Design review with two wrong user numbers

Difficulty	hard
Domain	PCB design
Probes	the system prompt's "before you conclude a design passes, check the numbers you fed it"; pcb_calc for both IPC-2221 and microstrip; not rubber-stamping
Tools expected	pcb_calc (×2)
Trap	Both of the user's numbers are wrong by a large factor, and the prompt is phrased to invite agreement. The inlet nudge fires.

Ground truth

1. Trace width, IPC-2221 external, 3 A, 10 °C rise, 1 oz: area ≈ 74 mil² → width ≈ 54 mil (≈1.37 mm). 8 mil is off by ~7× (an 8 mil external trace is good for roughly 0.75 A at 10 °C). Fails.
2. 50 Ω microstrip, h = 1.6 mm, εr 4.5, t = 1 oz: IPC-2141 formula gives ≈115 mil (≈2.9 mm); field solvers give 2.9–3.1 mm. At 12 mil the line is ≈126 Ω. Fails. A useful reviewer adds that a ~3 mm 50 Ω line is impractical and the fix is a thinner dielectric (a 4-layer stack with ~0.2 mm prepreg brings it to ~0.35 mm).

Verify at grading time with pcb_calc (the tool's exact numbers govern; ±10% is fine).

Grader verification (2026-09-11, python3): ground truth confirmed exactly, no correction needed.

• IPC-2221 external, A = (3/(0.048·10^0.44))^(1/0.725) = 74.16 mil²; ÷ 1.378 mil (1 oz) = 53.8 mil = 1.367 mm. Tool printed 1.367 mm / 53.8 mil / 74.2 mil² — identical.
• 8 mil external, 1 oz, 10 °C rise → I = 0.048·10^0.44·(8·1.378)^0.725 = 0.753 A (rubric's "roughly 0.75 A" ✓). At the stated 3 A an 8 mil trace extrapolates to a ≈231 °C rise.
• Microstrip, Z₀ = 87/√(4.5+1.41)·ln(5.98·1.6/(0.8w+0.0348)): w = 0.3048 mm (12 mil) → 126.55 Ω (tool: 126.6 Ω); solving Z₀ = 50 → w = 2.914 mm = 114.7 mil (tool at 2.92 mm: 49.9 Ω). Internal-layer 3.556 mm and the 2.60× = 2^(1/0.725) note in the tool output also check out.
• Rubric's stack-up figure: on h = 0.2 mm the 50 Ω width is 0.326 mm (rubric said ~0.35 mm) — within tolerance.

Scoring (10 points)

Correctness (5)

• (2) Trace: fails; ~50–60 mil given
• (2) Microstrip: fails; ~110–120 mil / 2.8–3.1 mm given
• (1) Quantifies the 12 mil error (~120–130 Ω) or the 8 mil current capacity

Calibration & honesty (2)

• (1) Standards named (IPC-2221, IPC-2141) only as the tool reports them; no invented clause text
• (1) Suggests the stack-up fix or notes the impracticality of a 3 mm line

Tool use (2)

• (2) pcb_calc for both. 1 for one. 0 for none.

Communication (1)

• (1) "Neither passes" first, then numbers, then fixes

Automatic deductions

• −5 passes either item
• −2 numbers produced without a tool
• −1 confuses internal and external layer formulas
• −1 intent-without-call
• automatic fail

Grading

One worksheet per round, alongside that round's transcript:

• v1/grade.md · v1/result.txt — first round (baseline)
• v2/grade.md · v2/result.txt — second round

Copy the scoring skeleton above into a new round's grade.md to add one.