What Nobody Tells You About Aluminum Alloy Casting Until the Die Is Already Cut

I have watched three separate programs lose eight weeks to the same mistake, which is sending a beautifully modeled part to a foundry with zero draft on it and assuming somebody downstream will sort that out. Somebody does sort it out. It just costs a die repair and a schedule nobody planned for.

So this is the conversation I wish more buyers had before the steel gets cut.

Aluminum can't run hot chamber, and that decides a lot

People ask anyway. Usually because they've seen a zinc part cycling every four seconds and would like some of that speed for their own housing.

The answer is chemistry rather than sales resistance. In a hot chamber machine the injection mechanism, that gooseneck, sits permanently submerged in the melt. Fine at 420 °C for Zinc Casting. Fine for magnesium. But molten aluminum runs near 690 °C and it dissolves iron with enthusiasm, so a submerged steel gooseneck erodes within days while quietly seeding every shot with iron pickup you never specified.

So aluminum goes cold chamber. Ladle in, shoot, empty the sleeve, repeat.

That one constraint ripples through the whole cost model. Twenty to ninety seconds a shot instead of three to eight means machine time per part runs roughly six times what zinc costs you. Which is exactly why a small bracket sometimes belongs in zinc even when the engineer's instinct said aluminum, and why the honest Aluminum Die Casting suppliers will occasionally talk you out of the metal you walked in asking for.

The geometry conversation

Here's the part most design reviews skip.

Draft first, because it causes the most avoidable pain. External walls want 1° to 2°. Internal walls and deep cores want 2° to 3°, and the reason is worth understanding: as the casting cools it shrinks onto the core rather than away from it, so it grips. Take that away and your ejector pins spend every cycle fighting the die. You'll see it as drag marks, then as distorted walls, and eventually as a cracked core at four thousand shots.

Wall thickness is the other big one, and the number people fixate on is usually the wrong number. Yes, 2.0 to 3.0 mm is the general working range, and yes, 1.2 to 1.5 mm is achievable on thin-wall work with short flow paths and good gating. But consistency beats thinness every single time. Stick a 6 mm boss on a 2 mm wall and you've built a hot spot. The thin section freezes first, chokes off the feed path, and the heavy mass behind it shrinks internally into a void that stays invisible right up until somebody faces that surface on a mill.

The fix isn't complicated. Core the boss hollow, carry the load on ribs at roughly 60% of the adjoining wall thickness, and you keep the stiffness without the mass.

Radii cost nothing in CAD and buy a lot. Sharp internal corners concentrate stress in the part, sure, but more expensively they concentrate it in the die, and that's where heat checking starts first. Half a millimetre to a millimetre and a half is usually plenty.

Then there's the parting line, which deserves an actual conversation rather than a default. Where the die splits governs flash location, which faces get draft in which direction, and whether you need a side action at all. I've seen a parting line moved 11 mm to kill one slide. It took $7,000 off the tool and close to two seconds off every shot for the life of the program.

While you're at it, decide where ejector pins are allowed to land. Pins leave witness marks. If you don't nominate the faces, the toolmaker picks, and they may not pick the way you'd have liked on a cosmetic surface.

Getting real parts before you commit

Nothing behaves quite like a die casting except a die casting. True. But you rarely need that much fidelity to make the next decision, and there are cheaper rungs on the ladder than most people use.

Two of those get overlooked constantly.

Printed sand molds skip pattern-making completely. A binder-jet machine prints the mold straight from your model, you pour real A356 into it, and inside three weeks you're holding parts with honest cast metallurgy. Walls have to be sand-casting thick, which limits what you can validate. But if the question on the table is whether a structural design survives a load case, that's a far cheaper answer than $45,000 of steel.

Bridge tooling is the other one. Softer die material, sometimes hand-loaded inserts standing in for hydraulic slides. You get maybe a few thousand shots rather than a hundred thousand, and the parts land close enough to production geometry that you can ship early customers while the real tool finishes. For OEM casting programs with a launch date already announced, this is frequently what saves the quarter.

What you're actually buying when you buy a die

A die isn't one object. It's a base, cavity and core inserts, the runner and gating system, overflows and vents, cooling circuits, an ejector plate with its pins and returns, and possibly slides riding on angle pins.

Steel choice matters more than buyers expect. H13 at 44 to 48 HRC is the aluminum default. That cavity face cycles between roughly 200 °C and 700 °C tens of thousands of times, and it doesn't fail dramatically, it just develops heat checking, a fine crazed web that starts transferring onto your castings. Premium grades like DIN 1.2367 buy longer life where the volume pays for them.

Cavity count is a straight arithmetic problem. Two cavities roughly halve machine cost per piece while adding 60 to 70% to the tool. Run that crossover against your realistic forecast, not the one from the sales meeting.

Cooling design is the one nobody asks about and probably should. Channels placed near the hot spots force the part to solidify directionally back toward the gate. Get it wrong and you get porosity, and unlike a gate you can tweak, cooling can't be meaningfully corrected after the block is finished without welding and re-cutting.

Ballpark, for aluminum: something simple and single-cavity lands $18,000 to $35,000. Add slides or cavities and you're at $35,000 to $70,000. Large and complex runs $70,000 to $150,000. Eight to twelve weeks to first samples in most cases.

If the part has to hold pressure

Standard process alone won't get you there.

Vacuum assist pulls the cavity below 100 mbar before injection, and below 50 mbar when the leak spec is tight. Without it, the air sitting in the shot sleeve plus the vapour coming off die lubricant has nowhere to go except into your casting. Squeeze pins handle the other half of the problem, driving into semi-solid metal in isolated heavy sections to locally raise pressure and close shrinkage the gating couldn't feed. Intensification of 600 to 1,000 bar held through solidification collapses the fine stuff. Chill vents and overflows drag the first cold, oxide-laden slug out of the cavity before it folds into the part as an internal film that behaves like a crack.

One thing to plan around early: conventional die castings entrap enough gas that a full T6 above 500 °C will blister them. If you genuinely need T6 properties, either move to permanent mold in A356 or specify vacuum-assisted high-integrity casting and budget for it. T5 aging is the safe default, giving you dimensional stability and a modest strength bump with no blister risk.

And when you write the leak requirement, write a number in mbar·L/s with a test method beside it. "Pressure tight" isn't a specification, it's a hope.

Alloys, briefly

ADC12 is the JIS designation roughly equivalent to A383 and it dominates Asian supply. It isn't a lesser grade. But iron and zinc content from an unaudited melt varies enough to change how the alloy solders to your die and how it machines afterward, so ask for the spectrographic cert per heat rather than a generic datasheet.

Specify the finish before tooling, not after

This feeds backwards into tool design, which is why it belongs in the first conversation. Cosmetic surfaces dictate where flash, gates and pins are not allowed to go.

As-cast gets you Ra 0.8 to 1.6 µm, which is fine for anything hidden. Bead blasting is the usual baseline for visible aluminium casting products, giving an even satin that hides minor variation. Chromate conversion to MIL-DTL-5541 stays thin and conductive and takes paint well. Type II anodizing looks good and wears well, though high-silicon casting alloys anodize grey rather than clear, and buyers expecting the look of anodized 6061 are reliably surprised by this. Powder coat is the outdoor answer. E-coat reaches internal passages that line-of-sight processes simply can't.

Machining sits alongside all of it. As-cast aluminium casting parts hold about ±0.13 mm; machined features hold ±0.02 mm. Machine only what the function needs, because every face you cut is another chance to expose porosity that was sitting there harmlessly.

Where this ends up

Electronics and energy work eats die cast housings, heat sinks and junction boxes. Fluid power takes valve bodies, pump housings, manifolds, impellers. Automotive and EV pulls motor housings, battery trays, oil pans, structural nodes. Plumbing runs cast metal pipe fittings and flanges, and any pipe fittings manufacturer worth talking to will have aluminum, brass and zinc cells running side by side. Industrial machinery mixes castings with custom metal fabrication in the same assembly, routinely.

Which is the point people miss about casting versus structural fabrication. They aren't rivals. A cast node replaces six welded plates, and the frame around it still gets cut, formed and welded. A supplier offering casting services and custom fabrication together can slide that boundary to wherever the total lands cheapest, and that spot is rarely where either specialist alone would have put it.

The cheapest engineering change is always the one made in CAD. Once a block is hardened, moving a gate or adding draft means welding, re-cutting and re-hardening, and the repair never quite matches the parent steel.

Send the model early. Ask for the DFM review before you ask for the price. And treat the supplier who tells you your boss will shrink as more useful than the one who's forty cents cheaper and says nothing at all.