# FR-4 mesh sensitivity

Separate solves at 2 W, 30°C, components upward. Effective package temperatures, not junction temperatures.

| In-plane spacing (mm) | Board layers | Cells | D1 (°C) | Q1 (°C) | R1 (°C) | U1 (°C) | Steady checks |
|---:|---:|---:|---:|---:|---:|---:|---|
| 1.00 | 4 | 128,760 | 179.14 | 109.95 | 129.31 | 97.20 | Passed |
| 0.75 | 6 | 236,430 | 172.59 | 113.06 | 125.33 | 98.56 | Passed |
| 0.50 | 8 | 570,752 | 170.45 | 112.83 | 124.39 | 99.18 | Passed |

The two finest meshes differ by:

- D1: 2.148 K (1.53% of the fine-grid temperature rise above ambient).
- Q1: 0.228 K (0.28% of the fine-grid temperature rise above ambient).
- R1: 0.944 K (1.00% of the fine-grid temperature rise above ambient).
- U1: 0.623 K (0.90% of the fine-grid temperature rise above ambient).

These differences measure sensitivity, not a certified error bound. Board-normal layer count, package/edge sampling, and resolved holes change with refinement; a simple uniform-grid Richardson extrapolation would be unjustified. No domain-size or radiation-angle convergence study, experimental validation, or aluminum-core mesh study has been completed. Steady convergence and spatial resolution are separate questions.