A 24 GHz Radar PCB: From Chip Reference Layout to Stackup Review

A 24 GHz reference layout is a starting electrical geometry, not a fabrication-independent drawing. Before ordering a Rogers RO4350B radar PCB, compare the chip vendor's reference construction with the board shop's proposed stackup. Close every difference that changes the RF fields, return path, connector transition or antenna environment. A purchase order that says only “RO4350B, controlled impedance” leaves too many decisions unresolved.

This article extends our 24 GHz radar PCB guide with a reference-to-production review worksheet. It addresses transferring a chip evaluation layout to a manufacturable board. It does not provide a validated antenna design or claim measured radar range.

Freeze one reference revision before comparing anything

For a concrete example, Infineon's EVAL BGT24LTR22 user guide, UG224433, revision 1.10, 15 April 2025, describes a four-metal-layer evaluation board. Section 3.1 specifies a 0.250 mm RO4350B RF core and says the matching structures are designed for that stackup. It also identifies blind vias beneath the SMA connectors. Section 3.3 defines device impedances at the PCB landing pads, looking into the pins. Section 4.3 distinguishes compensated interconnect loss from uncompensated matching-network and connector mismatch contributions. These are conditions of that reference, not guaranteed results for a new PCB.

Archive the exact manual revision, layout package and schematic used for the review. Give the fabricator the same revision identifier that the RF designer used. A similar chip name, an older evaluation board or a screenshot without units is insufficient. If the source drawing omits a manufacturing detail, mark it “unresolved”; do not silently fill it in from a different board.

The difference register: what must match, what must be re-evaluated

The following is an original engineering review worksheet. Its proposed changes are illustrative, not details of a supplier quotation. Assign an owner and a closure record to every row before fabrication.

A useful status system is “unchanged and evidenced,” “changed and validated,” or “unresolved.” Avoid “equivalent” without a measurable definition. The RF engineer approves electrical consequences; the fabricator confirms achievable construction. Neither signature substitutes for the other.

A small thickness change can consume the phase budget

Consider a deliberately simplified calculation, not a reconstruction of Infineon's matching network. Assume an isolated, uniform, uncoated microstrip with width w = 0.500 mm, relative permittivity εr = 3.66, zero conductor thickness, and no dispersion or loss. Compare dielectric heights h = 0.250 mm and 0.300 mm. The second height is a hypothetical substitution, not a recommended stackup.

For w/h > 1, use εeff ≈ (εr + 1)/2 + (εr − 1)/(2√(1 + 12h/w)), then Z0 ≈ 120π/[√εeff × (w/h + 1.393 + 0.667 ln(w/h + 1.444))]. At f = 24 GHz, λg = c/(f√εeff), with c = 299,792,458 m/s. Electrical phase over length L is 360L/λg degrees.

The predicted impedance shift is about 6 Ω even though the line artwork is unchanged. The approximately 3° phase difference over this illustrative 10 mm section is also relevant when allocating channel mismatch. These numbers are model outputs, not manufacturing tolerances or measured performance. Real matching discontinuities, coupled traces, soldermask, copper roughness, plated thickness and frequency-dependent dielectric behavior require a more complete model. Do not use this calculation to approve a 24 GHz layout.

The Rogers RO4000 laminate data sheet lists RO4350B design Dk of 3.66 for its stated design characterization and process Dk of 3.48 ± 0.05 at 10 GHz. Those values have different measurement contexts. Here 3.66 is a declared starting assumption, not an extracted value for a particular fabricated line. Ask which dielectric model and copper description the simulation and supplier use.

Preserve the chip reference plane, not just a “50 Ω” label

Split the RF route into named regions: chip landing pad, matching structure, nominal transmission line, launch or antenna feed. Record where a device model starts and where a fixture model ends. A board shop can meet a uniform-line impedance target while a changed pad, nearby ground cutout or connector produces a different input match.

For an antenna-integrated product, removing a coaxial connector is a design change. The antenna feed has its own impedance and environmental sensitivity. Establish the intended source/load planes and evaluate the new complete transition. If the reference includes four channels, preserve channel identities throughout layout, simulation and test reports so a swapped port or reversed coupon cannot masquerade as material variation.

Ask the fabricator for evidence that can be checked

1. Construction proposal: a layer-by-layer drawing identifying RF core, bonding layers, nominal and allowed dielectric heights, finished copper and via spans. Distinguish base foil from finished plated thickness.
2. Process capability: trace width/spacing control, registration, via geometry and the proposed measurement method. Request the relevant capability for this construction, not a generic marketing minimum.
3. Electrical coupon: a representative RF structure with documented geometry and reference planes. Decide whether the purpose is impedance, propagation phase, attenuation or transition characterization; one coupon does not automatically answer all four.
4. Acceptance definition: frequency range, port impedances, environmental conditions, calibration method, fixture removal and uncertainty. Fill in numerical limits from the system budget before placing the order.

Keep line loss, connector loss and antenna efficiency separate in the budget. For phase-sensitive channels, compare relative phase under the same measurement setup and define how cable movement, connector repeatability and temperature will be controlled. A supplier's bare-board impedance report is useful evidence, but it does not validate an assembled radar's detection range or angular accuracy.

Release the new board through three gates

• 1. Construction agreedEvery stackup difference has an owner and disposition.
• 2. RF model updatedChanged fields, transitions and antenna environment are represented.
• 3. Prototype correlatedCoupons and assembled-board tests use agreed reference planes.

Start with the lowest-cost prototype that can discriminate the competing explanations for a failure. If a uniform line fails, investigate construction and calibration before retuning the chip match. If lines correlate but the assembled input differs, inspect launch, assembly and device-plane assumptions. If conducted RF agrees but radar behavior changes in the enclosure, move attention to antenna loading, coupling and system processing.

RO4350B can be a useful candidate for this workflow; it is not the default answer for every low-frequency control section or every millimeter-wave radar. Use our radar material selection guide to frame alternatives and the millimeter-wave PCB overview when the target band changes. A successful 24 GHz reference transfer does not qualify a 77 GHz design.

Prepare a quote package that exposes the differences

For a useful manufacturing review, provide the chip/reference-board revision, operating band, array dimensions, channel phase targets, material candidates, proposed stackup, Gerbers or ODB++, drill/via details, copper and finish requirements, quantity, and test scope. Include the completed difference register and mark any material substitution as requiring engineering approval. State whether antenna, enclosure and assembly validation are inside or outside the requested work.

Source scope: Infineon UG224433 rev. 1.10, sections 3–4, and the Rogers RO4000 laminate data sheet. Review worksheet, example calculations and release workflow are original editorial guidance. No third-party figures reproduced; no simulated values are presented as laboratory measurements.