Copper is a dream for thermal conductivity and a nuisance to machine. C1100 (electrolytic tough-pitch) cuts almost like butter until the same softness that makes it easy leaves built-up edge on the flute, smears the wall, and throws a burr on every fin tip. On a telecom heat sink with 0.4 mm fins at 20,000 pieces, the job was never about removing metal — it was about keeping 0.4 mm walls standing straight and clean. Soft and gummy, copper rewards a sharp edge and a controlled cut and punishes a dull one. Get the tool and the feed right and the fins ship ready; get them wrong and 20,000 parts need a hand at the deburr bench.
Why copper smears in the first place
Copper's trouble is built into its crystal structure. It is face-centred cubic, ductile, and soft (about 40 HRB for C1100), so the chip welds to the cutting edge instead of breaking away. That weld is built-up edge, or BUE, and once it forms the tool is no longer cutting with its ground geometry — it is dragging a lump of stale copper across the wall. The wall smears, the finish goes grey and torn, and the next fin picks up a burr. Pure copper is worse than the alloys here; oxygen-free C101 is stickier than C1100, and tellurium- or lead-bearing free-cutting coppers help the cut but drop thermal conductivity by a few percent.
We pick C1100 for the 100% IACS-class conductivity and accept the BUE fight as part of the job. The fix is never more spindle power; it is a sharper edge, a higher shear rate at the tooth, and a tool that sheds the chip before it can weld. A heat sink like this is core to electronics machining — server lids, RF baseplates, and LED cold plates all lean on thin copper to pull heat off a die, so the surface the fin presents has to stay flat and clean, not smeared. Read the smear as a tooling signal, not a material defect, and the rest of the plan follows from it.
Sharp tools, no exceptions
C1100's gummy nature means a slightly worn edge builds material within the first few parts. We run polished-flute uncoated carbide with a sharp ground edge; on copper a mirror polish sheds the chip better than a TiAlN or AlTiN coating, which can grab. Coatings earn their keep on steel and Inconel, not here. Tools go to a fixed piece count, not to failure — on a 20,000-piece run a tool change at 800 parts costs minutes, while letting one run to 1,200 risks BUE on the last 200. We track edge wear under the microscope at the change interval, not by ear.
The first sign of BUE shows in the finish before it shows in the size, so finish is our early-warning gauge. We also run a positive rake angle and a high helix so the chip curls off instead of smearing along the wall. Edge prep matters as much as the coating does: a honed but still sharp edge clears better than a razor that chips on the first hard inclusion. We machine a wide range of materials beyond copper, but C1100 stays our most common heat-sink alloy because the conductivity is worth the fight, and the discipline to hold it transfers straight across to brass and aluminum too.
Feeds and speeds that shear instead of rub
Too slow a feed and the edge rubs instead of shears, which grows BUE and tears the finish. Too fast and the 0.4 mm fin deflects or chatters and the wall goes wavy. For C1100 we run a moderate surface speed — around 180–220 m/min on a polished carbide end mill — with a feed per tooth high enough to guarantee a shear cut rather than a rub. Climb milling keeps the tool biting into a thinning chip, which protects the fin wall better than conventional milling, where the tool tends to push the wall over on exit. We keep radial engagement shallow on the finish pass so the side load on the fin stays under a few newtons.
The numbers are not exotic; the discipline is. A 0.01 mm change in feed per tooth is the line between a clean wall and a smeared one, and we prove it on a first-article fin before the line runs. We log the exact surface speed, feed per tooth, and axial depth for every operation, because copper will happily cut at the wrong setting and only show the damage three shifts later. The cut that looks fine coming off the machine is the one that burrs at deburr — so we set the parameters by the burr they produce, not by the sound they make.
Keeping 0.4 mm fins upright
Thin copper fins deflect under their own cutting force, so the cut plan starts with the base. We rough the base and the fin roots with full stock, then leave the fins slightly oversize and take a light finish pass with minimal radial load so the wall barely flexes. Flood coolant at moderate pressure washes swarf out from between the fins; packed chips are how a fin gets knocked flat mid-cut, and a flat fin is scrap. The fixture clamps the base hard and leaves the thin top sections free to move — but they move only microns because the load is low.
On a 20,000-piece run the fixture is the real hero: a soft jaw that seats the base true and a locating pin that holds the part against the thrust, so every fin is cut from the same datum. We qualify the fixture on the first ten parts and re-check it every shift, because a 0.02 mm shift in the locate shows up as a 0.02 mm fin thickness drift across the whole batch. Copper is soft enough that a hard jaw will also dent the base, so the jaws are urethane-faced and torqued to a set value, not cranked down by feel. Hold the base, go light on the fins, and the walls stay where the program put them.
Burr control is the real deliverable
A heat sink with burrs on the fin tips will not seat against the device, and hand-deburring 20,000 parts is not a plan. The spec on this part called for no loose burr larger than 0.05 mm on the tips. We control the burr at the cut: sharp tools, a clean exit so the tooth does not tear the tip, and a light abrasive-flow or tumble pass sized to break the edge without bending a fin. On this part we used a soft-media tumble for 8 minutes per batch, which took the tips to a 0.02 mm max edge with no fin loss.
The recast and the smear from a dull tool are what create the big burrs, so burr control really starts with the tool plan two sections back, not at the deburr bench. Designing the part for the deburr step — a generous tip radius, no blind fin slots, enough pitch to get media between the fins — is what lets the process hold without touch-up. A sharp-tipped fin is the most expensive thing on the drawing; a 0.05 mm tip radius costs almost nothing in thermal terms and saves a hand-deburr line item on every part. We flag that trade on the quote, because the burr spec is where copper jobs win or lose.
Running 20,000 clean
Volume is where the small decisions compound. A 14-day run at 20,000 pieces means every tool change, every coolant top-up, and every fixture check has to happen on a schedule, not when something looks wrong. We set the tool life at a fixed count, log each change, and keep a shadow set of gauges on the fin thickness so a drift shows before it leaves the tolerance band. The fin thickness held 0.4 mm ±0.02 across the run, and fin height stayed within 0.05 mm of straight — no part needed rework.
Coolant concentration ran 8–10% at 22 °C, and we filtered to 25 µm so no grit scratched the soft walls. Chip evacuation got a dedicated nozzle per fin bank, because on 0.4 mm fins the difference between a washed slot and a packed one is a blocked line nobody noticed. The lesson we repeat on copper: the part is easy to cut and hard to cut clean, so the whole plan is built around the second half. A 20k run that looks fine at part 500 and burrs at part 15,000 is a run that was never in control, and control is what the schedule buys you.
Why it is cheaper than it looks
Copper costs more than aluminum per kilo, but the cycle is short because the metal removes fast, and the real cost driver — deburring — was designed out rather than added at the end. A shop that quotes copper like aluminum will either miss the burr problem or surprise you with a hand-work line item on 20,000 parts, which is the whole margin. We price by the feature and the deburr plan, not by the material, and we say so up front.
If your thermal design needs thin copper fins, bring the deburr spec to the quote and ask how the shop controls burrs at volume. The fins are easy to cut; they are hard to cut clean, and the clean part is what you are actually buying. A print that names the tip radius, the finish, and the burr limit tells us everything we need to burn the cost out before the run starts, and that is the conversation that keeps a 20k order from turning into a 20k rework. Send the drawing and we will tell you which callouts are doing real work and which are just decoration.
| Feature | Drawing requirement | Result held |
|---|---|---|
| Fin thickness | 0.4 mm ± 0.03 | ± 0.02 |
| Fin height | 8 mm | deflection ≤ 0.05 mm |
| Fin pitch | 1.2 mm | held, no packed chips |
| Tip burr | ≤ 0.05 mm | ≤ 0.02 mm, no hand work |
| Surface finish | Ra 1.6 µm | Ra 1.2 µm |
| Volume / lead | 20,000 pcs | 14-day run, 0 rework |
- Specify the fin tip radius you can accept — a 0.05 mm tip radius is far cheaper to deburr than a sharp one, and costs almost nothing thermally.
- Ask how burrs are controlled at volume; if the answer is hand work, walk away on a 20k run.
- Keep fin pitch open enough for coolant and a chip to clear; 1.2 mm pitch on 0.4 mm fins is the tight end.
- Fix the alloy to C1100 unless you need the slight smear resistance of a free-cutting grade, and say so on the print.
"On copper fins, a dull tool does not make a bad part — it makes a smeared, burred part that looks machined until you try to assemble it."
If your thermal design needs thin copper fins, design the deburr step in from the start and ask how the shop controls burrs at volume. The fins are easy to cut; they are hard to cut clean, and the clean part is what ships. Send the drawing and we will flag the callouts that matter before a single part is run.