Automotive 4D imaging radar is moving from 3-transmit / 4-receive to 8-transmit / 8-receive. What tightens is not only element pitch but antenna board area, hybrid stack-up layer count and lot-to-lot consistency. The more defensible procurement approach is to negotiate the antenna board, the main board and the test station separately, priced against material and process capability, instead of squeezing one integrated BOM.
1. Doubling the channels changes the board before it changes the algorithm
At 77–79 GHz the element pitch is set at half a wavelength, roughly 1.9 mm in air, and it tightens further once the wave is inside the laminate. An 8T8R aperture needs more than twice the elements and a longer feed network, which typically pushes the antenna board from 2 layers to 4. The main board then has to carry more transmit and receive channels plus additional power rails, so a 6–8 layer FR-4 hybrid with the antenna board has become the mainstream structure — and the hybrid interface has become the new reliability focus.
- Antenna board: 4 layers of high-frequency material, radiating elements and feed lines on one board.
- Main board: 6–8 layers of FR-4, hybridised with the antenna board or mated through a connector.
- Hybrid interface: the thermal expansion mismatch between high-frequency material and FR-4 can reach several times, making lamination and reflow the dominant failure zone.
2. Material choice: pay for Dk consistency, not a single Df number
On short feed lines, the insertion-loss gain from dropping Df from 0.0037 to 0.0015 is easily consumed by copper foil roughness. What actually drives lot yield is Dk variation between lots and dielectric thickness tolerance — a 0.05 Dk drift within one lot is enough to break phase consistency across an 8T8R array. Writing those two items into the technical requirement is more effective than chasing an attractive Df figure at the negotiating table.
- Hold Dk variation within ±0.05 per lot and tighten thickness tolerance one grade inside the laminate datasheet limit.
- High-frequency laminates commonly run 4–8 weeks lead time and ship in full sheets, so reserve volume against project milestones.
- Do not accept undeclared substitute grades; any material change requires fresh hybrid lamination and insertion-loss validation.
3. Move the tolerance upstream into the design
Phase error in a radiating array comes from feed-line length differences. At 77 GHz a 1 mm length difference is already on the order of a wavelength and cannot be recovered by tuning impedance. Designing equal-length feeds and opening the solder mask over the antenna area is cheaper than chasing ±8 μm line-width accuracy in production, and far easier to replicate across several board shops.
- Design feed-length differences within 0.1 mm and use radiused corners throughout.
- Open the mask over the antenna area or specify low-Dk solder mask so the overlay does not raise the effective Dk.
- Keep legend and test points out of the antenna area, and leave at least 3 mm from the board edge to the outermost element.
4. The production test station is often the longer wait
Millimetre-wave test needs fixtures, a chamber and calibration standards, and building that station routinely takes longer than the boards themselves. A buyer who watches only board-shop capacity will still stall at production test after the prototype has been validated. The safer route is to confirm the supplier's test capability during prototyping, including whether S21 data is supplied per lot and whether lot-to-lot comparison is possible.
- Require coupon-level S21 insertion-loss data with every lot; a certificate of conformity alone is not enough.
- Hold lot-to-lot insertion-loss shift within 0.3 dB and trigger material traceability beyond that.
- Write test-station count and throughput into the framework agreement rather than accepting a verbal commitment.
5. Five procurement calibration points
Risk in a radar programme sits at the material and test ends; price is the last item that should be discussed. The five points below can go straight into an RFQ template and also serve as a quarterly checklist against actual supplier delivery.
- Quote the antenna board and the main board separately so high-frequency laminate is not diluted by FR-4 pricing.
- Lock material grade and lot, and require written declaration plus revalidation for any substitute.
- First article to include microsection, hybrid-interface inspection and impedance coupons at ±8%.
- Keep a dual-source strategy of one primary plus one alternate, with hybrid lamination and insertion-loss re-test before the alternate runs production.
- Put material lead time, test-station capacity and throughput into the framework agreement and review them quarterly.
AIPCB.cc supports up to 44 layers, a maximum panel size of 610×1100 mm, an aspect ratio up to 20:1 (25:1 for backplanes), ±8% impedance tolerance (±10% standard), ±8 μm line-width accuracy and up to 13 OZ heavy copper, with 24-hour prototype turnaround. For high-frequency hybrid antenna and main-board evaluation, send your stack-up, material grade and insertion-loss targets to info@ipcba.com.cn, or call +86 138 2310 9230 to work directly with engineering.
Send your Gerber files and stackup requirements; a 24-hour expedite channel is available.