How wide is a 50 ohm trace on 1.6 mm FR-4?

About 3.02 mm as a surface microstrip with 1 oz copper, taking Dk 4.4 as the median of 22 FR-4 grades. Computed with the Hammerstad and Jensen equations. Confirm with your fabricator, because Dk varies between grades and with frequency.

How wide is a 50 ohm trace over a 0.2 mm dielectric?

About 0.35 mm on standard FR-4 with 1 oz copper. An etch tolerance of 0.02 mm moves it between 48.5 and 51.6 ohms, which is why thin dielectrics need a test coupon.

Why does etch tolerance matter for controlled impedance?

Impedance depends on the ratio of trace width to dielectric height. The same 0.02 mm etching error is a small fraction of a 3 mm trace and a large fraction of a 0.15 mm one, so fine traces over thin dielectrics swing several ohms on etching alone.

TOOL · DRAWN TO SCALE

How wide do I make the trace?

Pick the laminate by name, not by a dielectric constant you have to go and find. The cross-section below is the geometry that was solved, drawn to scale, and the result comes with what your fabricator's etch tolerance will do to it. 87 laminates from the catalogue.

Structure

LaminateDk 4.4 and Df 0.016, as published at 1 GHz, median of 22 grades.

Target impedance

Trace to plane, mm

Etch ± mm

Copper

0.353 mmtrace width · 13.9 mil

Impedance: 50.0 Ω
Effective Dk: 3.18
Delay: 151 ps/in
Loss at: 0.59 dB/in

An etch tolerance of ±0.02 mm moves this between 48.5 and 51.6 Ω, which is ±3.2%. Comfortable inside a ±10% specification.

Where the numbers come from

Microstrip uses Hammerstad and Jensen's 1980 equations with their strip-thickness correction. Stripline uses the exact conformal-mapping solution for a thin strip and Wadell's closed form for finished copper; our tests hold the closed form against the exact one, and they agree within 1.5 percent.

Differential pairs use the IPC-2141 coupling factors, which are approximations good to roughly ten percent. For a differential pair your fabricator's field solver is the authority: send them the target, the laminate and the layer, and let them return the geometry.

Dielectric constant is taken as published, at the frequency shown beside it. Real laminates move with frequency and resin content, so a finished board is confirmed with a test coupon, not a calculator.

See what each laminate does to a signal Choose a surface finish