F4B vs FR-4: Where the Frequency Limit Actually Falls

“F4B works to 40 GHz, FR-4 to about 3 GHz” is the sort of line that survives in blog posts and dies in a design review. Frequency limits are not a property of a laminate; they are the point where accumulated loss eats the budget on the length you actually need. This page does the f4b vs fr4 comparison the only way it survives scrutiny: dielectric loss per inch from 1 to 77 GHz, copper loss added in, and a 3 dB reach table that says how far each material runs at each band.

Ask a fabricator what the maximum frequency of FR-4 is and the honest answer is a question: how long is the trace and how many dB can you spend? A material does not stop working at a frequency; the loss per inch grows until the accumulated attenuation crosses the link budget. That is why the same FR-4 that carries a 30 GHz signal across 5 mm of a test coupon fails on a 150 mm feed line at 8 GHz.

So this comparison has three parts: the loss per inch each material produces, the copper loss that sits on top of it, and the reach that follows once you fix a budget. Everything else — band labels, adjectives, “ideal for mmWave” — is a summary of those three numbers.

For a 50 Ω line the dielectric attenuation is close to

α_d ≈ 2.6 · f_GHz · √Dk · Df   [dB/inch]

Two consequences follow immediately. Loss is linear in frequency, so a band twice as high costs twice the dB per inch. And loss is linear in Df, which is where the material choice lives: FR-4 at Df 0.020 produces roughly twenty-five times the dielectric loss of an F4B grade at Df 0.0010–0.0018. The √Dk term is the small one — moving from Dk 2.20 to 3.00 raises dielectric loss by only 17 % at equal Df, which is why the Df column, not the Dk column, decides the band.

FR-4 constants used here: Dk 4.30, Df 0.020 (mid-loss laminate, datasheet range 0.015–0.025). F4B grades use the Wangling official values — the Df ladder from 0.0010 to 0.0018 across the F4BM range is tabulated in the complete F4B model list.

Dielectric loss only. The ratio between the FR-4 row and any F4B row is roughly 27× at every frequency — a constant multiplier, because the equation is linear in f for both. Read this table alone and you would conclude the material swap is worth 14 dB per inch at 77 GHz, which is true and also misleading: the copper term is about to eat a fixed share of the budget, and on short traces it dominates.

For 1 oz copper at 50 Ω, conductor loss is about 0.19 dB/in at 10 GHz for FR-4 geometry and about 0.18 for the wider F4B lines — both scaling as √f. Add it in and the picture changes shape:

On FR-4, copper is a 36 % tax at 1 GHz and becomes irrelevant at 77 GHz because dielectric loss has taken over everything. On F4B the reverse holds: at 10 GHz the material contributes 0.039 dB/in and copper contributes 0.18, so the copper is now the larger term. This is the point most comparison articles miss, and it has a practical consequence — once the laminate is this good, copper roughness, finish and line width become the levers that move the number.

Fix a 3 dB attenuation budget — a common allocation for a feed network — and divide. The table says how much line each material can carry before the budget is spent:

This is the table to bring to a design review. FR-4 at 77 GHz supports 9 mm of 50 Ω line at 3 dB — less than a centimeter, which is why no radar front-end is built on it. The same FR-4 at 3.5 GHz supports 156 mm, which is why plenty of sub-6 hardware ships on FR-4 without apology.

There is a clean way to state the boundary: pick the trace length. On a 100 mm line with a 3 dB budget, FR-4 needs total loss ≤0.76 dB/in, and the arithmetic puts that crossover at about 5.7 GHz. Below it FR-4 is a legitimate choice; above it the same board needs a different laminate or a much shorter run.

Useful corollaries: a 30 mm interconnect on FR-4 clears 3 dB to roughly 10 GHz; a 300 mm cable assembly on FR-4 clears it only to about 1.9 GHz. The material did not change — the length did. That is also why the “FR-4 is fine below 1 GHz” claim survives: it is a safe statement for long backplanes, and far too strict for a short RF stub. The engineering reasons FR-4 still constrains high-frequency designs beyond loss are set out separately in why standard FR-4 is not suitable for high-frequency PCB design.

Loss decides whether the signal arrives; Dk tolerance decides whether it arrives in phase. FR-4 quotes Dk around 4.3 with a batch tolerance of roughly ±0.2, while F4B grades quote ±0.04–0.06 on their nominal value:

For a single serial link, 25° of batch-to-batch phase spread is annoying. For a phase-steered array, where elements must track within a few degrees, it is disqualifying — and that, not attenuation, is the real reason FR-4 disappears above the low bands. Impedance tolerance adds on top; the workflow for holding it is in the impedance control guide.

Grade choice inside F4B follows the same arithmetic: the 2.20 grade minimizes loss, the 2.55 grade trades 38 % more dielectric loss for 11 % smaller patches, and the 3.00 grades buy tolerance at 77 GHz. The middle rung is worked through in the F4BM255 positioning page, and the family ladder in the F4B family overview.

The page this replaces quoted several numbers that do not survive a unit check:

The calculation needs four inputs and takes minutes: band, longest 50 Ω run, dB budget, and the phase tolerance of the application. Run them against the reach table above and the material decision usually makes itself — and when it lands between two grades, the deciding question is tolerance, not loss.

XCE builds both sides of this boundary: FR-4 and high-Tg multilayer up to 72 layers, and PTFE and hybrid builds on 600×1200 mm panels with plasma or sodium-naphthalene desmear, in-house pressing and coupon verification, impedance ±10 % standard and ±3 % coupon-verified, samples in as fast as 24 hours, inspected to IPC-A-600G and IPC-6015, with 80+ engineers on 10,000+ builds since 2009. The material-side differences that follow a laminate swap are covered in high-frequency vs standard PCB processing; where the choice crosses into imported laminates, the Rogers materials guide holds the parameter anchors.