Where the program started
The customer sells commercial LED downlights and had a second-generation product selling well and fighting itself thermally. The luminaire used an extruded 6063 fin stack bonded with thermal adhesive to a die cast A380 housing carrying the driver, the gland and the lens interface. It is a common architecture, and it exists because extrusion makes cheap thin fins and casting makes cheap complex housings, so combining them looks like the best of both.
The problem is what sits between them. On production samples that bond line measured 0.26 K/W, so at the module’s 42 W dissipation, 11 °C of junction temperature was being spent on a joint whose only function is to connect two parts that did not need to be separate. Planning 60,000 units a year for the third generation, they asked whether the fins could be cast.
Why A413, and the conductivity argument in full
The honest starting position is that casting the fins makes the fins worse. Extruded 6063 runs around 200 W/m·K, while the best of the die casting alloys, A413, gives 121 W/m·K. Trading 200 for 121 looks like a step backwards.
It is not, because the two effects operate over completely different lengths. Fin conduction happens over 22 mm of height in a 2 mm section, and across 46 fins the difference between 200 and 121 W/m·K works out under 1 °C at 42 W. The interface being removed is worth 11 °C. Consolidation wins by roughly an order of magnitude, and it would still win with A380 at 96 W/m·K.
A413 was chosen anyway because there was no reason not to: highest conductivity available in high pressure die casting, excellent castability for thin fins, and a tensile strength of 296 MPa that is irrelevant on an unloaded part. The one real cost is machinability, rated fair because of the 11 to 13% silicon, which put the machined LED seat onto PCD tooling.
What cast fins can do that extruded fins cannot
Cast fins cannot be as thin as extruded fins. An extrusion will hold 1.0 to 1.2 mm; the cast fins here are 2.0 mm at the root tapering to 1.4 mm at the tip, with 2° draft per side and 1.5 mm root fillets, because a sharp fin root heat-checks the die at exactly that line.
What cast fins can do is be radial. An extrusion produces a constant cross-section along one axis, so its fins are parallel, and on a circular downlight that wastes the corners of the envelope and channels air one way. Cast fins radiate from the centre, taper along their height and vary in spacing around the circumference, which gave 14% more fin surface area inside the same Ø164 × 88 mm envelope. The 2° draft is not negotiable, so designing with that constraint from the start produced a better fin than fighting it would have.
What else came into the casting
Once the fins are cast, everything else may as well be too. The driver cavity, the M20 cable gland boss, four mounting bosses and the lens gasket land are all features of the single casting. The gland boss is the only feature that needed a slide.
The bill of materials went from two aluminum parts, an adhesive, four fasteners and an assembly operation to one part. That is the shape most consolidation savings take: the material saving is minor and the operations that stopped existing are the whole story.
Coating specified as a thermal parameter
The finish on a heat sink is not only cosmetic, and it is worth specifying with a thickness window rather than a minimum.
A 65 µm matte black polyester film raises surface emissivity from roughly 0.06 on bare blasted aluminum to about 0.90. On a natural-convection sink that is worth around 4 °C, while the film’s own conduction resistance at 65 µm costs under 0.5 °C. Above roughly 150 µm the balance reverses. One casting-specific step matters too: gas trapped in the casting expands during the 180 to 200 °C cure and escapes through the wet film, leaving pinholes in the fin roots, so every part is pre-baked at 210 °C for 25 minutes to degas before coating.
Where it landed
Measured junction temperature on the first production build was 81 °C at 42 W in a 40 °C ambient, against 92 °C for the bonded design and a CFD prediction of 83 °C. That 11 °C let the customer uprate the published L70 life from 50,000 to 70,000 hours, which was worth considerably more commercially than the cost saving.
Unit cost fell 31%, from $9.05 to $6.25, and the assembly operation with its 3.4 minutes per unit left the customer’s plant along with the visible seam between two coated substrates.
The part runs on a single-cavity tool at 60,000 units a year. At that volume a second cavity would pay back inside a year, which is a conversation worth having before the current tool reaches the end of its shot life rather than after.