Manufacturing

CNC Machining

/see-en-see muh-shee-ning/
CNC machining is the primary manufacturing process for precision RF passives: waveguides, cavity filters, diplexers, antenna feeds, and housings. Multi-axis mills (3 to 5 axis) cut complex 3D geometries in metals with tolerances of 5 to 25 μm and surface finishes of 0.4 to 1.6 μm Ra. A 10 μm error in a Ka-band waveguide (WR-28, 7.112 mm) shifts cutoff by ~5 MHz. Materials include aluminum 6061-T6, brass C36000, copper C110, and Invar 36 for thermal stability.
Category: Manufacturing
Tolerance: 5 to 25 μm
Surface finish: 0.4 to 1.6 μm Ra

Understanding CNC Machining for RF

RF and microwave passive components are among the most dimensionally demanding products manufactured by CNC machining. Unlike structural parts where tolerances serve mechanical fit requirements, RF component dimensions directly determine electromagnetic performance. A waveguide's internal width sets its cutoff frequency, a cavity filter's resonator dimensions determine center frequency and Q-factor, and a connector interface's concentricity affects return loss. Every micrometer of dimensional error translates to measurable degradation in insertion loss, return loss, isolation, or frequency accuracy.

The relationship between machining precision and frequency band creates a natural hierarchy. At L-band and S-band (1 to 4 GHz), waveguide dimensions are large (WR-284: 72 mm wide) and standard CNC tolerances of ±50 μm represent only 0.07% of the critical dimension. At Ka-band (26 to 40 GHz), WR-28 is only 7.112 mm wide, so the same ±50 μm tolerance becomes 0.7%, shifting the cutoff by 25 MHz and degrading return loss by several dB. At W-band (75 to 110 GHz), WR-10 is 2.54 mm wide, and even ±10 μm (0.4%) measurably affects performance. This progression drives the adoption of 5-axis CNC, precision grinding, and wire EDM for mmWave components, with correspondingly higher manufacturing costs.

CNC Machining and RF Performance

Frequency Shift from Dimensional Error:
Δf/f = -Δa/a   (waveguide cutoff, dominant mode)

Surface Roughness Loss Increase:
αrough = αsmooth × (1 + (2/π)arctan(1.4(Rq/δ)2))

Cavity Q Degradation:
1/Qactual = 1/Qideal + 1/Qmachining

Where a = waveguide broad dimension, Δa = dimensional error, Rq = RMS surface roughness, δ = skin depth. Example: WR-28 (a = 7.112 mm), Δa = 10 μm: Δf = -29.5 MHz at fc = 21.077 GHz.

RF Materials for CNC Machining

MaterialConductivityCTE (ppm/°C)MachinabilityRF Application
Aluminum 6061-T625 MS/m23.6GoodWaveguides, housings, antennas
Copper C11058 MS/m17.0FairHigh-Q cavities, waveguides
Brass C3600016 MS/m20.5ExcellentConnectors, adapters, small parts
Invar 361.0 MS/m1.2DifficultStable cavity filters (outdoor)
Kovar2.0 MS/m5.1DifficultHermetic packages, feedthroughs
Common Questions

Frequently Asked Questions

What tolerances are required for RF CNC machining?

Scales with frequency: ±50 μm at S-band, ±25 μm at X-band, ±10 μm at Ka-band, ±5 μm at W-band. Cavity filter tuning screws need ±25 μm positioning with 10 μm thread concentricity. Surface roughness above 0.4 μm Ra at W-band increases loss by 0.01 to 0.05 dB per interface (skin depth only 250 nm at 100 GHz).

What materials are used for RF components?

Al 6061-T6: general purpose (25 MS/m conductivity, light, corrosion resistant). Copper C110: highest conductivity (58 MS/m), 30 to 50% lower conductor loss. Brass C36000: best machinability, used for connectors. Invar 36: CTE 1.2 ppm/°C for thermally stable cavity filters (10 to 30 ppm drift vs 200 to 500 ppm in aluminum). Kovar for hermetic seals.

When is 5-axis CNC needed for RF parts?

Required for curved surfaces (corrugated horns), compound angles (orthomode transducers), and undercuts (internal coupling irises). Eliminates re-fixturing errors (10 to 25 μm per re-fixture). Essential above 60 GHz where total tolerance budgets are under 25 μm, making single-setup 5-axis the only way to meet specifications.

Precision RF Manufacturing

Request a Quote

Need CNC-machined waveguides, cavity filters, or custom RF housings in aluminum, copper, or Invar? Contact our team.

Get in Touch