How CNC Machining Solves the Base Station Heatsink Crisis
Massive MIMO radios push 200 W and more into an enclosure the size of a briefcase. Air alone cannot move that heat. This guide shows how CNC machining solves the fin, base and interface problems that castings and extrusions leave behind, and where the process is the wrong answer.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
Why CNC Machining Solves Heat Flux That Casting Cannot
A modern macro cell packs 64 or 128 transceiver chains into one box. Each chain has a power amplifier running at 30 to 50 percent efficiency, so a 200 W radio can dump 100 W or more as heat inside a sealed aluminium shell. The die area under each amplifier may be only 20 × 20 mm, which puts local heat flux above 100 W/cm². Total watts matter less than that local number.
Air cooling works by moving heat from the die into the base, along the base, then out through fins. Every step adds resistance. If the base cannot spread heat sideways fast enough, the junction runs hot no matter how large the fin stack is. This is where the manufacturing method starts to matter more than the fan curve.
Extrusions give you one constant cross-section. A cast fin needs 1.5 to 2° of draft on each side or it will not release from the mold, and thick fins at the root to survive the pour. Both limit how much surface area you can fit into a fixed enclosure volume. CNC machining solves that by cutting fins from solid, so the wall stays vertical and the root stays thin.
The practical result: a machined heatsink can carry 0.8 to 1.2 mm fins at 2.5 to 4 mm pitch, with base thickness of 6 to 12 mm tapered toward the edges. That geometry is not castable and not extrudable, but a 5-axis machine cuts it without special tooling.
- 1Draft-free finsVertical walls, no release angle needed.
- 2Thin rootsMore fins in the same footprint.
- 3Local thickeningExtra material only under hot components.
Material and Interface Choices That Decide the Outcome
Aluminium 6061-T6 is the default. Thermal conductivity sits near 167 W/m·K, it machines cleanly, and hardcoat anodizing adds a hard, corrosion-resistant skin. For higher flux, 6063 and 6082 also machine well. Copper C110 runs at roughly 390 W/m·K, about 2.3 times aluminium, but it weighs 3.3 times as much and costs far more per kilogram.
That trade favors inserts. A copper coin or block set into an aluminium base puts high conductivity only under the hot die, while the rest of the heatsink stays light. Bonding methods matter: a press fit alone leaves a cold interface. Soldering, brazing or a thermal epoxy with a thin bond line of 0.05 to 0.15 mm keeps the joint from becoming the new bottleneck.
The base-to-heat-spreader interface is where most teams lose performance for free. A milled surface at Ra 1.6–3.2 μm with 0.1 mm flatness can leave 30 to 50 μm of air gaps once the two parts are bolted. Air conducts at about 0.026 W/m·K, so those gaps dominate. Machining the base to 0.02 mm flatness over 100 mm and Ra 0.8–1.6 μm, then using a 0.1 mm thermal pad or a thin grease layer, removes most of that resistance.
Stainless and titanium are the wrong choice for the thermal path. They are useful for brackets, screws and RF shields, but 304 stainless conducts at about 16 W/m·K. Use them where strength or corrosion matters, not where heat has to travel.
- 16061-T6General purpose base and fins, anodizes well.
- 2C110 copperInsert or vapor chamber housing under hot dies.
- 36063 / 6082Good finish, slightly lower strength.
Fin Geometry and Tolerance Limits Worth Paying For
Aspect ratio is the first constraint. For a 1 mm fin, keep height below about 20 to 25 mm, which is a 20:1 to 25:1 ratio. Push past that and the tool deflects, the fin leans, and vibration during cutting gets worse. Tall, thin fins also bend during handling and shipping, so account for that in the pack design.
Fin pitch sets how much air can pass. Natural convection needs 8 to 12 mm between fins or the boundary layers merge and the extra area does nothing. Forced air with a blower can use 2.5 to 4 mm pitch. Decide the cooling mode before you fix the pitch, not after.
Tolerance is where money is spent or saved. Fin thickness and pitch can usually live at ±0.1 mm. Fin height and overall length are fine at ±0.2 mm. The surfaces that carry heat need the tight work: base flatness of 0.02 mm over 100 mm, and mounting hole position within ±0.05 mm so the board lands without stress. Reserve ±0.005 mm for bearing bores and RF connector seats, not for every dimension on the drawing.
Surface finish follows the same logic. Ra 0.8–1.6 μm on the thermal interfaces, Ra 1.6–3.2 μm on fins and outer faces. A polished fin does not reject meaningfully more heat at these temperatures, so the extra cost buys nothing.
- 1Keep 20:1 aspect ratioBelow 25:1 for 1 mm fins.
- 2Match pitch to airflow8–12 mm passive, 2.5–4 mm forced.
- 3Tight only where it countsBase and mounting, not fin flanks.
How to Check the Design Before Cutting Metal
Run a thermal simulation with the real interface resistances, not ideal contacts. Set the base-to-spreader gap at 30 μm and the bond line at 0.1 mm. If the result is marginal in the model, it will fail on the bench. Most teams find that improving flatness buys more than adding fins, because it removes a resistance rather than adding surface area.
Ask for a DFM review before the first cut. We return a quotation and free DFM analysis within 12 hours, and production can start within 24 hours after approval. For a heatsink, that review should flag fins thinner than 0.8 mm, aspect ratios above 25:1, and any thermal face that carries a cosmetic finish requirement.
On the first parts, measure the things that were designed to be tight: flatness, fin thickness, pitch, hole position. A 0.02 mm flatness value on the drawing is only useful if someone checks it. We inspect 100 percent of parts before shipment and provide reports on request.
Then run the real thermal test with the actual radio board and the actual airflow. Simulation tells you where to look. The bench tells you whether it works.
- 1Model with real gaps30 μm at the base joint, 0.1 mm bond line.
- 2Request DFM earlyCheaper than redesigning after the first cut.
- 3Test with the real boardAirflow and die layout both change the answer.
Step by Step: From Thermal Model to Machined Heatsink
- 11. Fix the heat map firstList every heat source with its wattage and die footprint. Mark anything above 20 W/cm². That map decides where the base gets thick and where copper inserts go. Skipping this step is the most common reason a heatsink misses its target.
- 22. Choose cooling mode and fin pitchPassive: 8–12 mm pitch, fins 15–25 mm tall, vertical orientation. Forced air: 2.5–4 mm pitch, fins 8–15 mm tall. Write the airflow direction on the drawing so the fins are not run crosswise to the duct.
- 33. Set base thickness and taperStart at 8 mm under the hottest die, 6 mm elsewhere, tapering to 4 mm at the edges. Thicker bases spread heat better but add mass and cost. A 2 mm increase in base thickness can add 15 percent to part weight.
- 44. Design the fin root and cornersUse a 0.3 to 0.5 mm root fillet. Sharp internal corners break tools and trap chips. Leave 1 mm of clearance at fin ends for the cutter to exit cleanly.
- 55. Pick the process routeOne to roughly 500 pieces: machine from solid plate. 500 to 5,000: machine from a near-net casting or extrusion, then cut fins and interfaces. Above 5,000: die cast with a machined base. The thermal interface is machined in all three routes.
- 66. Specify tolerance and finish per surfaceBase and mounting faces: 0.02 mm flatness over 100 mm, Ra 0.8–1.6 μm. Fins and cosmetic faces: ±0.2 mm, Ra 1.6–3.2 μm. Connector and bearing seats only: ±0.005 mm.
- 77. Add finishing for the environmentHardcoat anodizing for outdoor and coastal sites, clear anodizing indoors, electroless nickel where the part also needs EMI or wear resistance. Mask thermal interfaces if the coating would add more than 20 μm.
- 88. Inspect before it shipsCheck flatness on a granite surface or CMM, measure fin thickness at three heights per fin, and verify weight. A 5 percent weight error usually means a wall or base thickness drifted.
Which Route Fits Which Volume and Flux
| Route | Typical volume | Best flux range | Watch out for |
|---|---|---|---|
| Machined from solid plate | 1–500 pcs | High, above 50 W/cm² | Material cost per kg |
| Machined near-net casting | 500–5,000 pcs | Medium to high | Casting porosity under thin fins |
| Die cast + machined base | 5,000+ pcs | Low to medium | Draft limits fin density |
| Extrusion + machined base | Any, if profile fits | Low, uniform load | One cross-section only |
| Skived fins + machined base | 500–10,000 pcs | Medium | Fin height below 40 mm |
| Copper insert in aluminium base | Prototype to 5,000 | Very high, local | Bond line thickness |
Frequently Asked Questions
Can you machine a heatsink from copper instead of aluminium?
Yes, C101, C103 and C110 copper all machine well, and copper runs at about 390 W/m·K against 167 W/m·K for 6061-T6. The trade is weight and cost: copper is roughly 3.3 times denser, so a solid copper heatsink is much heavier than the aluminium part it replaces.
For most base stations, a copper insert of 20 to 40 mm diameter under the hot amplifiers gives most of the benefit at a fraction of the weight. We machine the pocket to a press or braze fit and keep the bond line at 0.05 to 0.15 mm.
What flatness do you actually hold on a heatsink base?
We hold 0.02 mm flatness over 100 mm on thermal interface faces as a standard callout, with a general tolerance of ±0.005 mm where the drawing requires it. Fin thickness and pitch usually sit at ±0.1 mm, which is enough for airflow performance.
The number that matters is the one on your drawing. If the base is called out at 0.1 mm, you get 0.1 mm. Tightening it to 0.02 mm is a machining change, not a tooling change.
When is die casting cheaper than CNC machining?
Above roughly 5,000 units per year, a die casting with a machined thermal interface is usually the lower-cost route. The mold cost is spread across the volume, and the casting carries most of the shape.
Below that volume, the mold cost dominates. Between 500 and 5,000 pieces, machining from a near-net casting or extrusion often gives the best balance: the casting provides the bulk, and CNC cuts the fins and the flat interfaces.
How thin can the fins be?
We machine fins down to 0.8 mm reliably at 20:1 aspect ratio, and 1.0 to 1.2 mm at higher ratios. Below 0.8 mm the tool pressure deflects the fin and vibration marks appear on the flanks.
Thin fins also bend in handling. If the heatsink ships as a loose part, add a protective comb or keep the fins at 1.2 mm minimum. Fin pitch matters as much as fin thickness: passive designs need 8 to 12 mm, forced air can use 2.5 to 4 mm.
Does anodizing hurt thermal performance?
A thin clear or hardcoat anodized layer of 10 to 25 μm adds little resistance and helps corrosion and emissivity. Thicker coatings and powder coats do add a measurable barrier, so we mask thermal interface faces when the coating would exceed 20 μm.
For outdoor and coastal installations, hardcoat anodizing is the usual choice. For indoor units, clear anodizing or a bare machined interface works and costs less.
What information do you need for a heatsink quote?
Send the 3D model or a 2D drawing with the thermal faces marked, the expected volume, the heat load per component, and the cooling mode. Material and finish preferences help, but we can recommend both.
We return a quotation and free DFM analysis within 12 hours. Uploads stay confidential, and we sign an NDA on request when the design is sensitive.
Send Us Your Heatsink Model
Upload a STEP file and we will return a quotation with free DFM analysis within 12 hours, and flag any fin or interface that will not hold tolerance before the first cut.
12-hour quote and DFM±0.005 mm toleranceRa 0.8–1.6 μm100% inspection
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
