Differential Pair Skew and Length Matching Calculator
Enter the data rate, rise time and your length mismatch. The calculator gives the skew in picoseconds, checks it against both the unit interval and the edge, and separates the part a serpentine can fix from the part it cannot.
| Skew source | ps | % of UI | Note |
|---|---|---|---|
| Length mismatch | the only one you control with a serpentine | ||
| Fibre weave over the pair | routing angle and laminate weave style | ||
| Total | added directly, not in quadrature |
Runs entirely in your browser. Nothing is uploaded, stored, or sent anywhere.
What skew actually costs you
Intra-pair skew — one half of a differential pair arriving before the other — does two separate kinds of damage, and they have different thresholds.
The first is eye closure. The receiver sees the difference of two signals; if one is late, the crossing moves and the eye narrows by roughly the skew. Measured against the unit interval, a few percent is usually tolerable and that is what a length-matching rule is protecting.
The second is worse and less often budgeted: skew converts differential signal into common mode. A perfectly balanced pair radiates very little because the two halves cancel. Skew breaks that cancellation, and the common-mode component it creates is what leaves the board on cables and shows up in an EMC chamber. That mechanism is governed by the rise time, not the bit period, so a slow interface with fast edges can have an EMC problem while its eye looks perfect.
tp ≈ 85 · √Dk,eff ps/inch ·
skew = mismatch × tp ·
UI = 1 / data rateMicrostrip and stripline differ here. A stripline is fully embedded and sees the laminate’s full Dk; a microstrip has field lines partly in air, so its effective Dk is lower and it is faster — typically 140–150 ps/inch against 170–180 for stripline. That also means the same physical mismatch costs less time on an outer layer, and that a pair which changes layers accumulates skew unless both halves change together.
Fibre weave: the skew a serpentine cannot fix
Laminate is glass cloth in resin, and the two have quite different dielectric constants. A trace running directly over a glass bundle propagates at a different speed from one running over the resin between bundles. On a differential pair with a fixed separation, one half can sit over glass while the other sits over resin for a long distance — producing skew that has nothing to do with length.
This is why the tool separates the two sources. Length mismatch is deterministic and a serpentine removes it. Weave skew is a property of where your traces happened to land relative to the cloth, it varies panel to panel and even board to board, and no amount of length matching corrects it. The fixes are different in kind: route at a small angle to the weave so both halves average over glass and resin, specify a spread-glass or mechanically-spread style, or use a tighter-woven laminate. On a marginal design this term can exceed the length mismatch you spent board area serpentining away.
Serpentines are not free
- The tight bends in a serpentine are impedance discontinuities. Keep the amplitude modest and the spacing at least three times the trace width, or you trade skew for return loss.
- Match near the source of the mismatch, not at the far end. Skew that exists for most of the route has already converted to common mode before you correct it.
- Over-matching is real. Chasing the last few mil on a pair whose weave skew is several times larger is wasted effort and wasted board area.
- Inter-pair matching (between pairs, for source-synchronous buses) is a different and usually looser budget than intra-pair. Do not apply an intra-pair rule to a whole bus by reflex.
More free tools
Each of these runs entirely in your browser. Nothing is uploaded, stored or sent anywhere, and none of them needs an email address.