High-frequency laminates: Rogers, Taconic, F4B
Above a few hundred megahertz the laminate stops being a passive substrate. Its dielectric constant sets how wide a 50-ohm trace has to be, and its loss tangent decides how much of the signal reaches the far end.
- MaterialsRogers, Taconic, F4B
- Layers1 – 32
- Impedance controlListed process
- Min line/space0.06 / 0.06 mm
When FR-4 stops being good enough
Above a few hundred megahertz the laminate stops being a passive substrate and becomes part of the circuit.
FR-4 has a dielectric constant around 4.2 to 4.6 and a loss tangent around 0.02, and both vary from batch to batch. At 100 MHz that is invisible. Above roughly a gigahertz two things happen: the wavelength gets short enough that a 10% spread in Dk moves a supposedly 50-ohm trace outside spec, and the loss tangent turns into an amount of signal that measurably never arrives at the far end.
High-frequency laminates are sold on those two numbers. PTFE-based materials run Dk around 2.2 to 3.5 with loss tangents in the 0.001 to 0.004 range; hydrocarbon and thermoset products sit between them and FR-4. Those are ranges from the laminate suppliers' datasheets, not measurements of our process — and the sheet for the exact grade you buy is the one the board has to be designed against.
Which is the ordering constraint people get wrong: the laminate has to be picked before the stackup, because the trace widths follow from its Dk. A Gerber drawn on FR-4 assumptions and then moved to Rogers is a re-layout, not a material substitution.
What the datasheet numbers actually decide
Four figures on a laminate datasheet, and what each one costs you if it is off.
| Datasheet figure | What it decides | What happens when it is wrong |
|---|---|---|
| Dk — dielectric constant | How wide a 50-ohm trace has to be on a given dielectric thickness. | A 10% error in Dk moves the impedance by several ohms. On a tightly controlled net that is a failing coupon, and a re-spin. |
| Df — loss tangent | How much signal is absorbed per unit length. | It accumulates with length, so a short board can hide a bad choice that a long one exposes. This is the number that separates FR-4 from PTFE laminate. |
| Dk tolerance and batch consistency | Whether the impedance holds across production, not only on the first article. | It is the difference between a design that yields and one that needs impedance coupons on every lot. |
| CTE and Tg | Whether the vias and the BGAs survive reflow and thermal cycling. | A laminate that is right electrically can still crack a via barrel, especially on a thick board with a high layer count. |
Copper roughness is the one that surprises people: at high frequency the current crowds to the surface, so a foil with a rough profile has a longer electrical path than a smooth one, and the difference shows up as loss rather than as impedance. It is a property of the copper on the laminate, which is why two suppliers quoting "the same" Dk can still deliver different loss.
The stackup is the quote
We do not publish a single impedance tolerance, because it is not a property of the plant — it is a property of the stackup you send.
Trace width, dielectric thickness, the Dk tolerance of the laminate and the copper roughness all move the impedance together. A stackup with a tight Dk and a well-controlled dielectric can hold a few percent; the same plant with a loose laminate and a thick dielectric cannot, whatever the process claims. So the drawing fixes it: which layers carry controlled impedance, what the target is, and what the tolerance has to be.
Send four things with the Gerber — the stackup, the nets that need control, the impedance target and tolerance, and the finish. Impedance control is on our process list; ENIG, immersion silver and OSP are the finishes that suit a high-frequency board, and all three are on the finish list. The finish comparison covers when each one is the wrong choice.
High-frequency fabrication limits
The processing side is the same as the rest of the plant. What the laminate changes is the stackup, not the drill or the finish list.
| Base material | FR4, CEM1, CEM3, High TG, Rogers, F4B, Taconic, FR1, FR2, 94V0, aluminium |
| Layer count | 1 – 32 layers |
| Minimum line width / spacing | 0.06 mm / 0.06 mm |
| Impedance control | On the special-process list |
| Board thickness | 0.2 – 8.0 mm ± 10% |
| Copper weight | 0.5 oz – 12 oz |
| Maximum board dimensions | 650 mm × 1,100 mm (25.6 in × 43.3 in) |
| Minimum drill hole diameter | 0.1 mm |
| Surface finish | ENIG, immersion silver and OSP are the ones a high-frequency board usually takes; HASL, immersion tin, ENEPIG, electrolytic hard gold, gold fingers and ENIG + OSP are also available |
| Special processes | Half holes, bonding, impedance control, metal cladding, resin plug holes, blue glue, carbon ink, blind routing, blind and buried vias |
| Quality standard | IPC-A-600F / MIL-STD-105D |
| Order quantity | 1 – 500,000 |
Where a high-frequency board fits, and where it does not
A high-frequency laminate is right if…
You want a different family if…
If the design also needs a 0.4 mm-pitch BGA fanned out, the routing goes with HDI rather than with the laminate choice. Compare the five board types.
Same plant, other families: FR-4 and high-TG rigid · HDI build-up · flexible and rigid-flex · aluminium and heavy copper.
Common questions
Do you build boards on Rogers and Taconic laminate?
Yes. Rogers, Taconic and F4B are on our base-material list, alongside FR4, high TG, CEM1, CEM3, FR1, FR2, 94V0 and aluminium, from 1 to 32 layers.
Can you hold a controlled impedance on FR-4?
Yes — impedance control is on our process list and FR-4 supports it at lower frequencies. What FR-4 cannot do is hold the tolerance at multi-gigahertz edge rates, because its dielectric constant varies too much between batches.
What impedance tolerance can you hold?
That is set by the stackup rather than published as a single number: trace width, dielectric thickness, Dk tolerance and copper roughness all move it together. Send the stackup and the nets that need control, and the tolerance comes back with the quote.
Which surface finish should a high-frequency board use?
ENIG, immersion silver or OSP — all three are on our finish list and all three suit a high-frequency board. HASL's uneven surface is the one to avoid once the pitch gets fine.
How many layers can a high-frequency board have?
One to 32, the same range as everything else we fabricate. What changes with the laminate is the stackup, not the layer count.
The laminate is chosen before the stackup
A Gerber drawn on FR-4 assumptions and moved to Rogers is a re-layout, not a material substitution. Send the stackup and the nets that need control, and the trace widths come back fixed to the laminate you buy.