The cooling section in our Thorium had to dissipate 150 W of heat from a small, sealed industrial enclosure. That's a lot of power in a small envelope, and getting it out the right way turned out to involve more decisions than the brief suggested. Here's what stood out to me from the months we spent on the design.
Key learnings
ΔT is the metric, not absolute temperature. People outside thermal design usually ask "how cool does it run?" That's the wrong frame. The number that matters is the temperature rise (ΔT) between the inside of the enclosure and the ambient air at a defined heat load. Ambient changes; the cooling design is fixed. With industrial electronics tolerating ~85 °C continuously and worst-case field ambient sitting around 60 °C, the design budget for ΔT lands at about 25 K. That single number drove almost every decision downstream.
Manufacturing sets the ceiling, not physics. The geometry that's optimal in pure heat-transfer terms is usually highly curved, sometimes hollow, with internal features no end mill can reach. Given enough budget, you could chase the physical ideal. In a production part, the practical limit is whether the geometry can be made affordably at volume. The work isn't to design against manufacturing constraints; it's to design the best geometry within them.
Material and finish are thermal decisions in disguise. Aluminium 6061 for the thermal-conductivity-to-machinability ratio. Glass-bead blast for surface emissivity and a uniform matte finish. Blue anodise for corrosion resistance without disturbing the thermal path. None of those choices are cosmetic, and getting any of them wrong shows up later as either a thermal penalty or a field-reliability problem.
Why we didn't go with metal 3D printing
The obvious question for a heatsink like this is whether to print it. Additive manufacturing in metal lets you build geometries CNC can't reach: internal lattices, organic flow paths, fin shapes that approach the heat-transfer optimum. We wanted that to be the answer, and did the analysis to find out.
It wasn't, for two reasons. The cost-per-part gap for finished aluminium AM is still considerably larger than the milled equivalent at our production volumes. And the practical performance delta, measured as actual ΔT reduction at the operating point, wasn't large enough to flip the cost case. The 25-K budget could be hit with milled geometry; the marginal degrees of headroom from additive weren't worth the unit-cost hit.
We're not done with the technology. Cost-per-part is improving, and the secondary processing (heat treatment, surface finishing on AM aluminium) is maturing. We'll look at it again when the economics get closer, or when a product variant demands a tighter thermal budget than subtractive can deliver.
My takeaway
A heatsink is one of those parts that looks unassuming on a spec sheet – a chunk of milled aluminium, three small fans, a fin pattern. The actual work was in the decisions behind it. Pick the metric first. Pick the manufacturing process second. Shape the geometry within those constraints rather than against them. The thermal calculation itself was the easier half.
We're shipping the first units with this cooling design now. ΔT stays inside the 25-K budget across the operating range. What I'm watching next is the long-term thermal behaviour in real field deployments. That's where we'll find out whether the design holds up beyond the lab.