LED thermal lab

Troy's horizontal results at 30°C: FR-4 D1/Q1/R1/U1 = 120.8 / 107.6 / 109.2 / 94.6°C; metal core = 97.1 / 98.0 / 95.3 / 94.5°C. His setup used compact package models that are not fully specified in the archive. The new solid-envelope temperatures are therefore a historical comparison, not an apples-to-apples junction-temperature validation.

The STEP board was 0.410 mm thick; fabrication records specify 1.6 mm. This model uses 1.6 mm; the actual solved Autodesk geometry is unavailable to confirm whether Troy made that correction. Thickness can materially change the hotspot. Board properties and the 2 W heat allocation come from the recovered original setup; metal-core dielectric properties, package conductivities, and emissivity remain assumptions.

• Three-dimensional natural convection, ideal-gas buoyancy, Sutherland air transport, anisotropic board conduction, diffuse-gray radiation, and solid/air thermal coupling.
• A graded orthogonal mesh with stair-stepped board edges and package envelopes. Small holes and features may be unresolved at coarse spacing. Inspect mesh refinement before drawing conclusions from small differences.
• No fan, enclosure, cable heat sinking, resolved copper traces, solder fillets, or calibrated semiconductor junction network. FR-4 uses Troy's effective conductivity tensor. Metal core uses a nominal series/parallel layer model.
• Steady numerical gates require passing mesh checks, a heat imbalance below 1%, less than 0.1 K peak drift over 200 iterations, mass imbalance below 0.1%, peak speed drift below 1% over 200 iterations, finite velocity, and physical air/solid temperature bounds. Passing these does not establish mesh independence or experimental accuracy.
• Iteration is solver progress, not seconds of physical warm-up. The velocity arrows show direction only; hover on the color field for the actual speed.

Source audit and validation requirements · Editable operating point and assumptions · Mesh sensitivity report · Numerical results (JSON)