RF PCB Material Qualification: A Construction Selection Worksheet

RF ENGINEERING / MATERIAL QUALIFICATION

RF PCB material construction selection means approving a complete layer construction against a circuit requirement. A useful comparison identifies the exact laminate, dielectric thickness, foil, bonding layers, electrical model and acceptance evidence. A brand name or a matching Dk number is insufficient. This worksheet turns a multi-brand shortlist into a documented decision that a designer, purchaser and fabricator can review together.

The scope is a new design or a proposed material change before production release. It applies to thermoset and PTFE constructions, including mixed RF and digital boards. It does not establish interchangeable materials or certify a supplier. For an introduction to material families and early design tools, start with the RF PCB material selection guide; this article concentrates on the evidence needed to close a construction decision.

1. Write the requirement before comparing datasheets

Create one record for each critical signal path. Identify its operating band, routed length, reference planes, launches and any layer transitions. Separate allowable transmission loss from return loss and phase error: a line can satisfy one limit while failing another. State whether the product must work at room temperature only or across temperature, humidity and assembly exposure. Put an owner beside every requirement that is still provisional.

Use three evidence labels throughout the record: documented for an applicable supplier value, assumed for a design input awaiting confirmation, and verified for a result obtained under the agreed test conditions. A blank entry means unknown, not zero. Keep typical values distinct from procurement limits, even when both appear on a document headed “specification.”

2. Normalize the evidence, not just the units

Moving values into a common spreadsheet does not make their test methods equivalent. Preserve the source frequency, temperature, specimen thickness, direction and measurement method beside each Dk or Df entry. Distinguish a process-control Dk from the Dk intended for circuit modeling. If a datasheet does not specify a condition, write “not stated in this source” and request clarification instead of supplying a convenient assumption.

The following manufacturer examples illustrate three different reasons to inspect the construction. They are not a performance ranking or an approved substitution list.

• Isola Astra MT77: the published construction table lists a 0.0100-inch core at Dk 3.00 and Df 0.0017 at 10 GHz. A 1067-glass prepreg entry with 77% resin and 0.0032-inch thickness lists 2.91 and 0.0018 at the same frequency. Core data cannot silently become the prepreg model. See the manufacturer construction table.
• Rogers RO4350B: the standard table reports process Dk 3.48 ± 0.05 at 10 GHz/23°C by clamped stripline, while its footnote specifies 3.33 ± 0.05 for the 4-mil laminate. The table separately reports design Dk 3.66 using differential phase length over 8–40 GHz. The thin-core exception and the value’s intended use both matter. Consult the RO4000 datasheet, page 3.
• Shengyi SG7350D2: its product page identifies a glass-reinforced PTFE/ceramic construction. Typical process Dk 3.55 and process Df 0.0020 are given at 10 GHz/23°C using IPC-TM-650 2.5.5.5, with typical values based on a 0.508 mm specimen. That is evidence for a specific reported condition, not an arbitrary thickness or a measured 77 GHz board. See the SG7350D2 product data and remarks.

Apply the same discipline to Panasonic, AGC/Taconic, Arlon, Wangling and other candidates: obtain the exact grade and construction documentation before entering numbers. Leave a candidate on hold if only a family brochure is available. A shorter shortlist with traceable evidence is more useful than a large table containing unqualified precision.

3. Translate the system allowance into a line requirement

Illustrative allocation, not measured board performance: suppose a 100 mm interconnect has a maximum insertion-loss allowance of 1.50 dB at 10 GHz between defined external reference planes. Reserve 0.40 dB for the two launches together, 0.20 dB for transitions, and 0.20 dB as engineering margin. The remaining line allowance is:

L_line = 1.50 − 0.40 − 0.20 − 0.20 = 0.70 dB
α_allow = L_line / length = 0.70 / 100 = 0.007 dB/mm

Here L is a positive loss allocation in dB, and α is an average allowable line loss per millimetre at the stated frequency. For a well-matched passive path, insertion loss is commonly expressed as −20 log₁₀|S₂₁|. The subtraction is an early budget, not a complete cascade calculation: significant mismatch and interactions between discontinuities require network or electromagnetic analysis.

The 0.70 dB allowance includes conductor and dielectric contributions along the line; it is not a Df limit. Neither a datasheet Df nor a zero-roughness calculation establishes compliance. Evaluate the intended foil, geometry, dielectric and surface treatment together. If the line becomes 150 mm while all other allowances stay fixed, the limit becomes approximately 0.00467 dB/mm. Route length has changed the decision without any change in operating frequency.

Use the insertion-loss calculator to explore its stated model, then copy the assumptions into your record. Check the sourced Dk/Df comparison table for research starting points. Neither tool replaces construction-specific modeling or the coupon acceptance plan.

4. Turn candidate names into buildable alternatives

For each candidate, ask the fabricator for a proposed layer sequence before freezing the artwork. List the RF dielectric and its adjacent reference plane explicitly. Treat core and pressed bonding thickness separately. Confirm which dimensions the fabricator may adjust to meet impedance, and which dimensions belong to a resonator, antenna or component launch and must remain under design control.

Run two clearly labeled comparisons when considering a replacement. First hold the existing artwork constant to expose the risk of substitution. Then allow a redesign to the same electrical targets to determine whether the candidate could be viable. Do not compare an optimized candidate against an unchanged original and attribute the whole difference to the laminate.

Record fabrication questions as open actions: compatible bonding materials, treatment of drilled holes, copper balance, dimensional compensation, finished copper and inspection access. A hybrid stackup review is especially useful when an RF outer layer shares a board with digital or power layers. A familiar brand combination is not evidence that every layer order, thickness and press sequence will work.

Ask for a quotation against the same drawing revision, quantity, foil and test scope for every candidate. Separate material availability from electrical suitability, and include the cost of coupons, fixtures and qualification. Keep supply or schedule assumptions dated. A proposed saving has little meaning if the alternative omits the evidence required for release.

5. Design a coupon plan that can answer the decision

Include structures representative of the production layer, foil, finish and routing direction. A useful starting set is a short through connection, a longer straight line and a representative transition. Select lengths and calibration structures for the measurement method and usable frequency range; do not reuse arbitrary coupon dimensions without checking their suitability. An impedance-only test cannot establish insertion-loss compliance.

Agree which reference planes define the board requirement. If the acceptance limit includes launches, retain them in the reported path. If the goal is line-only characterization, specify an appropriate calibration or validated fixture-removal method and retain its supporting data. Keysight’s fixture and de-embedding guidance explains why fixture effects and reference planes must be handled explicitly. A difference between two raw traces is not automatically a validated material-loss measurement.

1. Before fabrication: approve the coupon drawing, construction, frequency sweep, port reference impedance, limits, measurement conditions and report format.
2. At measurement: identify board and lot, record calibration, fixture and instrument settings, inspect connector repeatability, and retain complex S-parameter data rather than only a screenshot.
3. At review: compare relevant band edges as well as the nominal frequency. Assess loss, match and phase separately, with measurement uncertainty and repeatability considered against the remaining margin.
4. Before release: define the number of specimens and lots appropriate to product risk. Resolve unexplained differences between model and measurement before calling the construction qualified.

If environmental behavior matters, repeat the agreed checks after the specified exposure or at the relevant operating conditions. Document stabilization and conditioning. A room-temperature first article does not establish lifetime reliability or performance over the full temperature range.

6. Release the construction and control later changes

Use three outcomes: release when the required evidence is complete; conditional prototype when a named uncertainty needs a defined experiment; or hold when a required limit or document is missing. Each outcome should identify who approves it and what would reopen the decision. This prevents “please quote this material” from becoming accidental production approval.

The manufacturing handoff should contain the board revision and files, a layer-by-layer material schedule, dielectric and copper requirements, controlled geometry, substitution restrictions and the accepted test plan. Link each report to that same revision. Changes in foil, bonding construction, dielectric thickness or process should trigger an assessment of the affected model and evidence, even if the laminate brand stays the same.

Sources and engineering scope

Sources checked 7 October 2026. Rogers: RO4000 datasheet, revision code 1592 080322, publication 92-004. Isola: Astra MT77 online product data, typical-value revision D (April 2021) and construction-table revision C (May 2024). Shengyi: SG7350D2 online product data; the page does not display a document revision. Confirm controlled supplier documents for procurement. The linked Keysight guidance supports measurement context; the worksheet, loss allocation and diagrams are original explanatory examples.