Thermal Agent

Say what heats your part, how it is cooled and how hot it may get. The agent solves it on the real shape, judges it against your limit and — if it runs too hot — tries fixes by itself (material, cooling, fins), solving every one.

Tip: say the surface finish — “black anodised”, “anodised”, “painted”, “bare” or “emissivity 0.85”. Not said = bare metal (low emissivity, the cautious choice).
Chip: name the package (“TO-220”, “TO-247”, “QFN”, “CPU”, “LED COB”) or its size (“die 10 x 10 mm”), “theta jc 0.5 K/W”, “thermal paste / pad / solder” and “Tj max 150 C” to get the junction temperature.
Chip on a circuit board (no heat sink needed): describe the board (“80 x 60 mm two-layer board”, “4-layer”, “2 oz”, “copper pour 25 x 25 mm”, “ground plane”, “9 thermal vias”, “vertical board”), the package (“SO-8”, “SOT-23”, “QFN 5 x 5”, “D2PAK”) and its datasheet “theta ja 35 K/W” (or “theta jb” + “theta jc top”) for the best accuracy.

1 · Your words become a heat problem: heat in (watts), cooling (still air, fan, water or your h value), limit.
2 · Solved on the real 3-D shape with our verified heat solver, refined until the temperatures settle.
3 · Verdict: PASS (5 °C margin), MARGINAL or FAIL.
4 · If needed it tries one change at a time — then combinations — solving each, prefers a design our real air-flow check can run on, and confirms the best with a full solve.

Heat-sink recipe within 0.15 °C of classic fin-array theory; cooling fin within 0.02% of fin theory; heat in = heat out exactly.

Quick answer: in still air the cooling strength is a realistic estimate for each design (fin-channel correlations at the solved surface temperature, 13% below them for fins up to 80 mm long and up to 28% below them for fins 250 mm or longer, matched to our real air-flow runs for short fins and to an independent OpenFOAM check for long fins); with a fan it is computed from the fan and your sink (the fan's flow against the sink's pressure drop and a channel model checked against published measurements within 2.5% and an independent OpenFOAM model, on the cautious side; typical fan values unless you give yours; never below still air; a fan on top blowing down is not computed yet); with water it is a handbook value. "Confirm with real air flow" computes it with our lattice-Boltzmann flow solver (still air only, no fans; sink standing upright; small heat sinks with fins at least about 2 mm thick and gaps at least 3 times the fin thickness; one run at a time for all visitors). Heat is spread over the named face, unless you name the chip: then it enters only through the chip's contact patch (spreading in the base solved, checked against the closed-form spreading resistance within 0.2% and OpenFOAM within 0.5%) and the junction temperature adds the interface material and the junction-to-case resistance (typical datasheet values unless you give yours; the real air-flow check does not cover chip answers yet). A chip on a circuit board is computed with our board model (copper layers and FR4, natural convection and radiation from both faces of the board, still air, one chip in the middle of the board, +5% safety margin; checked against published measurements within 6% and an independent OpenFOAM model within 4%); the answer shows a likely range: ±15% with your datasheet theta ja, ±25% with theta jb + theta jc, ±30% with typical package values. Steady state. In still air the heat radiated to the room is included (emissivity from the surface finish you name; none named = bare metal, 0.05, the cautious choice; between fins the surfaces mostly see each other, so only the channel openings and outer faces count; checked against published measurements and an independent OpenFOAM model); with a fan too. With water, or an h you give, radiation is not added.