Which PCB stackup tool actually matches fab results?
Published 2026-11-13
Use a 2D field solver that accepts your fabricator’s exact materials and press‑out data; Polar Si8000/Si9000, Speedstack with solver, and the solvers in Altium Layer Stack Manager and Cadence consistently correlate to coupon TDR, while equation‑only tools (IPC‑2141/2221, Saturn) are good for first estimates but miss roughness, mask, and anisotropy effects.
Field solver vs formula: a quantified example
Take an outer microstrip over FR‑4 with Dk = 3.7 at 1 GHz, copper t = 35 µm, dielectric H = 0.20 mm, and target Z0 ≈ 50 Ω. With W ≈ 0.32 mm and no solder mask, an IPC‑2141‑style equation typically predicts ≈ 50–51 Ω. A field solver using the same geometry plus measured copper roughness (Huray/Hammerstad) often lands ≈ 0.5–1.5 Ω lower. Add 20 µm of LPI mask (Dk ≈ 3.1–3.5) and the same trace typically measures another ≈ 1–2 Ω lower on TDR. Net effect: a nominal “50 Ω” can read ≈ 48–49 Ω unless roughness and mask are modeled.
Which inputs dominate impedance?
Frequency‑dependent Dk/Df from the laminate data sheet drives both Z0 and loss; generic FR‑4 spans Dk ≈ 3.6–4.2 and Df ≈ 0.012–0.020 at 1 GHz, while low‑loss FR‑4 runs Dk ≈ 3.4–3.7, Df ≈ 0.006–0.012. Press‑out per prepreg ply sets H: common values are roughly 7628 ≈ 0.18–0.20 mm, 2116 ≈ 0.10–0.12 mm, 1080 ≈ 0.06–0.08 mm after lamination. Copper thickness matters twice on outer layers because plating adds about 8–12 µm to the foil. Copper roughness lowers impedance and raises loss; if your solver supports it, add vendor Rz/Ra with a Huray or Hammerstad model. Solder mask over microstrip (≈ 15–25 µm) commonly reduces Z0 by about 1–3 Ω.
Example 4‑layer and 6‑layer geometries
For a 1.6 mm 4‑layer with controlled outer microstrip, a practical build uses L1 1 oz, L2 0.5 oz plane, a ≈ 0.9–1.0 mm core to L3 (0.5 oz), and L4 1 oz, with 2116+1080 prepregs giving ≈ 0.18–0.22 mm to each plane. With Dk = 3.7 and mask off, 50 Ω microstrip at H = 0.20 mm and t = 35 µm typically needs W ≈ 0.30–0.33 mm; with mask on, reduce W by ≈ 0.02–0.03 mm. For 6‑layer inner striplines targeting tight coupling, with t = 18 µm copper and H ≈ 0.10–0.12 mm to each plane, 50 Ω single‑ended often lands at W ≈ 0.17–0.20 mm; 90 Ω differential starts around W ≈ 0.10–0.12 mm and S ≈ 0.12–0.18 mm. Always re‑solve with your fabricator’s press‑out and mask status before release.
How do I hit 50 Ω and 90/100 Ω every time?
Work from the fabricator’s laminate codes, glass styles, and foil weights inside the solver, not generic FR‑4. Keep outer‑layer H near 0.15–0.22 mm so 50 Ω microstrips avoid extreme widths. Use symmetric prepreg pairs to keep striplines centered after press. For ≤ 0.10 mm lines, prefer inner stripline with ≈ 0.08–0.12 mm dielectrics rather than pushing ultra‑narrow/wide microstrips. Set impedance tolerances up front; ±10% is common, and ±8% is achievable with tighter material control and coupon feedback.
What to include in your RFQ stackup file
- Target impedances and tolerance (e.g., 50 Ω single‑ended, 90/100 Ω differential, ±8–10%)
- Layer order and target copper weights (specify base plus expected outer‑layer plating)
- Acceptable laminate families and required Dk/Df at the test frequency (or vendor codes)
- Preferred prepreg glass styles (e.g., 2116/1080) and any no‑substitution notes
- Solder‑mask presence/color on controlled‑impedance layers
- Test: coupon TDR per IPC‑TM‑650 2.5.5.7; acceptance per IPC‑6012 Class 2 or Class 3
Verification and acceptance with your fabricator
Request coupon‑based TDR per IPC‑TM‑650 2.5.5.7 on each controlled geometry. Ask for the as‑built stackup report listing measured press‑out by ply, foil thicknesses (including outer‑layer plating), mask thickness, and TDR results. If outer‑layer microstrip is mask‑sensitive in your design, place both masked and unmasked coupons.
When should I move to RF/high‑frequency laminates?
As serial rates push 10–14 Gb/s (Nyquist ≈ 5–7 GHz) and above, FR‑4’s Df (≈ 0.012–0.020) drives dielectric loss that can reach ≈ 0.6–1.2 dB/in at 10 GHz for 50 Ω stripline, depending on copper roughness and geometry. Low‑loss materials (Df ≈ 0.003–0.008) often cut that to ≈ 0.2–0.5 dB/in. Differential skew from glass weave also grows with tighter pairs; if you must hold 100 Ω within ±5% on fine geometry, use spread‑weave or controlled glass mixes (e.g., 2116 + 3313), keep pairs tightly coupled over planes, and solve with frequency‑dependent Dk/Df and a roughness model.
Acceptance, compliance, and deliverables
Boards can be built and inspected to IPC Class 2 or Class 3 acceptance criteria, and RoHS/REACH compliance is supported. For procurement, include these requirements in the PO so test coupons and reports are delivered with the lot.
Send your target impedances and laminates—HyperPCB returns a quote in 24 hours and DFM feedback in 1 business day. — Get a quote