CNC Machining of Base Station Radiator: How Machining Precision Sets Thermal Performance

• 1
  Where the heat actually goes in a base station radiator
• 2
  Which tolerances on a base station radiator housing actually earn their cost
• 3
  Machined versus die-cast: choosing the right base station radiator construction
• 4
  How the machining process shapes fin quality and flatness
• 5
  Measuring a base station radiator before it leaves the shop
• 6
  Machined billet versus die casting versus extruded radiator bodies

A remote radio unit puts 20 W to 80 W of waste heat into a housing the size of a shoebox. That heat leaves through three resistances in series: the die-to-case joint, the case-to-radiator interface, and the radiator-to-air surface. Machining only touches the middle and the last of those. It cannot fix a bad thermal pad, and it cannot fix blocked airflow.

The interface is where most field failures start. A machined base with 0.05 mm of bow leaves a gap that thermal grease cannot bridge. Air is a poor conductor, roughly 0.026 W/m·K, so a 0.1 mm air pocket under a 50 mm × 50 mm footprint adds about 0.8 K/W. At 40 W that is a 32 °C rise that never appears in the simulation.

Fin geometry sets the air side. Thinner fins and tighter pitch increase surface area, but they also increase boundary-layer interference and pressure drop. For natural convection in still air, 2 mm to 3 mm fin thickness with 8 mm to 12 mm pitch is a practical band. Forced convection with a fan or blower can push pitch down to 4 mm to 6 mm.

Wall thickness matters too. A die-cast housing may run 3 mm to 4 mm walls; a machined base can go to 1.5 mm where the load path allows. Thinner walls cut weight and thermal mass, which helps transient response but reduces the buffer during a short overload.

• 1
  Interface flatnessThe single most common cause of hot spots in returned units.
• 2
  Fin pitchMatch it to the cooling mode, not to what looks dense.
• 3
  Base thicknessThicker spreads heat sideways; thinner saves mass.

Not every dimension on the drawing deserves ±0.005 mm. Tightening a fin tip radius to ±0.005 mm costs cycle time and buys nothing thermally. Tightening the mounting face that touches the power amplifier does buy something. Sort the drawing into functional surfaces and clearance surfaces before you send it out.

Functional surfaces: the die or PCB mounting face, the waveguide or filter mating face, and any RF ground plane. These usually need flatness in the 0.02 mm to 0.05 mm range over the contact area, plus surface finish around Ra 0.8–1.6 μm so a thermal pad or grease layer wets evenly.

Clearance surfaces: outer walls, fin roots, cable routing features, and cosmetic edges. General machining tolerance is fine here. Holding ±0.005 mm on a fin root adds cost and can even cause chatter on tall thin fins.

Threaded holes and press-fit bores sit in between. A stripped M4 thread in a 6 mm aluminium wall is a field failure. We usually specify thread depth of 1.5× to 2× diameter and keep at least 1.5 mm of wall around the hole. On 6061-T6 that holds up to repeated assembly better than 6063.

• 1
  Mounting faceFlatness 0.02–0.05 mm, Ra 0.8–1.6 μm.
• 2
  RF mating faceTighter flatness, watch for burrs on the sealing groove.
• 3
  Cosmetic surfacesGeneral tolerance; do not pay for precision you cannot measure.

Die casting wins on unit cost at volume. Aluminium ADC12 flows well into thin fins and the tool pays back somewhere in the low thousands of parts. The catch is porosity. A casting can trap gas near the base, and that gas pocket sits directly under the hottest component.

CNC machining of a base station radiator from 6061-T6 or 6063 billet gives you a dense, porosity-free base. Thermal conductivity lands around 167 W/m·K for 6061-T6 and about 200 W/m·K for 6063, both better than typical ADC12 near 92 W/m·K to 96 W/m·K. That difference matters most in the spreading direction, not in the fins.

The practical answer for many programs is a hybrid. Cast the finned body, then machine the mounting face, the RF interface, and the threaded holes. This keeps tool cost low and puts precision only where the heat and the signal need it.

Fully machined housings make sense for prototypes, low-volume or custom sector configurations, and any unit that has to pass a strict thermal test with no porosity allowance. A 5-axis machine can cut fins, pockets, and angled faces in one setup, which removes the stacked-tolerance problem you get from flipping the part between operations.

• 1
  Billet 6061-T6Dense base, good strength, easy to anodize.
• 2
  Billet 6063Higher conductivity, softer, better for pure thermal blocks.
• 3
  ADC12 castingLow unit cost at volume, check porosity under the die.
• 4
  HybridCast body plus machined critical faces, often the best trade.

Tall thin fins vibrate. A 40 mm tall fin at 2 mm thickness deflects under cutting load, and the result is a tapered fin with a rough flank that both reduces area and looks bad under inspection. The fix is not always a slower feed. Reduce the unsupported length, use a smaller radial engagement, and support the fin with a sacrificial web if the design allows.

Base flatness after machining is not the same as base flatness after finishing. Anodizing builds 5 μm to 25 μm per surface depending on the type, and hardcoat can add more. If the mounting face is anodized, the flatness you measured on the machine is not the flatness the customer gets. Mask critical faces or plan a final lapping pass after coating.

Heat treat and stress relief belong in the plan, not in the recovery. A 400 mm long housing machined from extruded stock will move when material is removed. Rough machine, stress relieve, then finish. Skipping that step shows up as a bowed base that fails flatness inspection three days later.

Deburring is a thermal issue, not just a safety issue. A burr bridging two fins blocks airflow and creates a hot stripe. Automated tumbling and brushing work on external edges; internal fin valleys usually need hand work or a controlled media process.

• 1
  Rough then finishLeave 0.3–0.5 mm stock, stress relieve, then take the final pass.
• 2
  Fin supportUse webs or reduce depth-to-thickness ratio above 15:1.
• 3
  Mask before coatingProtect mounting faces if flatness is called out.

Flatness on a large machined face is easy to measure badly. A dial indicator on a surface plate gives you a reading at one point; a coordinate measuring machine with a scanning head gives you the whole map. For contact faces above 100 mm square, ask for the map, not a single number.

Fin pitch and thickness are best checked with an optical comparator or a vision system. Calipers on a 2 mm fin will compress the burr and read optimistically. If the drawing calls out pitch tolerance, the inspection method should match that tolerance.

Thermal performance is verified by a test, not by a drawing. A simple setup with a heater block and thermocouples on the base and at the fin tips will show whether the interface is making contact. A 5 °C spread across the base under load usually means the mounting face is not flat or the fastener torque pattern is uneven.

Records matter for repeat orders. We keep inspection reports on file and can supply raw material certificates, in-process records, and final dimensional reports on request. For telecom programs that run for years, that traceability is what keeps batch 20 matching batch 1.

• 1
  CMM scanUse it for any contact face over 100 mm square.
• 2
  Optical checkBetter than calipers for fin pitch and thickness.
• 3
  Thermal testHeater block plus thermocouples beats a paper curve.