Millimeter-wave radar antennas live or die on dimensional repeatability after etch, press, and plating. Teams that treat 24 GHz and 77 GHz as the same laminate shopping list usually discover the gap in the chamber, not in the BOM spreadsheet.
At 77 GHz (typically 76-81 GHz automotive bands), wavelength shrinks so that a few microns of undercut or a few percent of dielectric thickness drift move resonance and sidelobes enough to fail production calibration. Many 24 GHz short-range modules still have room for pragmatic material and etch windows; copying those notes into a 77 GHz RFQ is a common China-fab quoting failure mode.
This piece compares what actually changes for overseas buyers: loss budget, copper roughness, hybrid stackups, and the fab notes that keep antenna performance correlated from prototype to volume.
Why the industry shifted bands
Wider available bandwidth at 77 GHz improves range resolution and target separation. The shorter wavelength also shrinks antenna apertures for a given gain target, which helps corner radar and imaging-radar packaging.
ADAS functions such as adaptive cruise, automatic emergency braking, blind-spot monitoring, and imaging MIMO arrays benefit from that resolution. Industrial and traffic radar followed for similar reasons. The PCB implication is not marketing: phase-sensitive multi-channel feeds need stable Dk, controlled copper profile, and etch geometry that a digital FR-4 note set will not guarantee.
What changes electrically
Dielectric loss (Df) and conductor loss both rise in importance as frequency climbs. Copper foil roughness becomes a first-order insertion-loss term because skin effect concentrates current at the surface. Dk stability over temperature and humidity matters as much as the published room-temperature Dk used in early line calculators.
Antenna element length, gap, and feed geometry inherit those material realities. A laminate that "looks low loss" on a brand page can still fail if thickness control, glass weave, or foil type drifts between lots.
| Focus | 24 GHz radar | 77 GHz radar |
|---|---|---|
| Loss budget | Often manageable with careful routing | Df and copper roughness dominate |
| Etch / feature tolerance | Important | Typically much tighter |
| Stackup / thickness control | Needed | Critical for resonance |
| Hybrid FR-4 + RF layers | Sometimes practical | Useful but process-risky |
| Coupons / chamber correlation | Recommended | Strongly recommended |
Material properties that belong in fab notes
Call out preferred laminate part numbers and allowed alternates for RF layers. Silent CAM substitutions are a frequent cause of identical Gerbers producing different chamber results.
Document copper foil type or roughness class on antenna and feed layers, not only "1 oz copper." Specify finished dielectric height targets and where thickness is RF-critical versus mechanical-only. Environmental expectations (automotive soak, industrial humidity) belong in the same package so the fab does not optimize only for room-temperature coupons.
Manufacturability is part of selection: lamination cycles, drill smear behavior, registration, and yield at the etch window you need. The lowest Df option that the shop cannot hold is not the lowest-risk option.
Hybrid stackups and routing geometry
Not every layer needs premium RF laminate. Many modules keep antenna or feed layers on low-loss material and place digital, power, and control sections on more economical constructions. That hybrid approach saves cost but changes press cycles, CTE balance, and registration risk - treat it as an early DFM topic.
Transmission-line choice (microstrip, stripline, grounded coplanar waveguide) must match the stackup you can actually buy. Grounded coplanar waveguide is common on compact 77 GHz feeds because it helps control fields in dense modules, but it still needs consistent reference and etch fidelity.
Process controls buyers should demand
Ask how etch tolerance is measured on antenna-critical features, what registration budget applies between RF copper and reference planes, and whether microsections or RF coupons ship with first articles. Vague "tight etch" language produces uneven quotes and uneven builds.
If cavities, windows, or selective plating appear on the drawing, put them in fab notes rather than Slack threads. Prior spin lessons - humidity detune, vendor foil swaps, chamber fail root causes - save weeks on the next release.
Prototype correlation before volume
Simulation alone does not lock production material and etch windows. Define what first articles must prove: feed S-parameters, chamber pattern checks, or both. Align fab coupons with the same stackup you will buy in volume.
When prototypes pass and volume fails, uncontrolled process drift that never entered the notes is a usual root cause. Freeze frequency-specific laminate PNs, antenna-layer etch limits, hybrid layer maps, and coupon expectations before the next RFQ.
For related process context see high frequency PCB, impedance control PCB, and automobile electronics PCBs. Send radar stackup packages through how to place an order when you want a DFM pass against real process limits.