Near-end and far-end crosstalk from your geometry, plotted against spacing with the 3W point marked. It separates the two because they behave completely differently: NEXT saturates with length, FEXT does not — and stripline has no far-end crosstalk at all.
| Spacing | S/H | NEXT | FEXT | Total (both sides) | vs budget |
|---|
Runs entirely in your browser. Nothing is uploaded, stored, or sent anywhere.
This is the single most useful fact about crosstalk and it is missing from most calculators. Far-end crosstalk is the difference between two coupling mechanisms: inductive coupling, which induces a voltage of one polarity, and capacitive coupling, which induces the opposite. In a homogeneous dielectric those two terms are exactly equal, and FEXT cancels to zero.
A stripline is fully surrounded by laminate, so it is homogeneous, so it has no FEXT. A microstrip has field lines partly in laminate and partly in air, the two couplings no longer match, and the residue is far-end crosstalk that grows with every inch of parallel run.
The practical consequence is large: if far-end crosstalk is your problem, burying the bus is not an incremental improvement, it is the fix. Switch the layer type above and watch the FEXT column vanish.
NEXT = Kb · V once the coupled length exceeds
Lsat = tr / (2 · tpd), and proportional to length below itFEXT ∝ Kf · V · Td / tr —
linear in length, inverse in rise time, no ceiling
Near-end crosstalk propagates back towards the driver. Once the coupled section is longer than half the rise-time length, the returning energy overlaps itself and the amplitude stops growing: NEXT is saturated, and making the parallel run longer does not make it worse.
Far-end crosstalk travels forward alongside the signal, accumulating the whole way. It has no saturation length. So on a short link NEXT dominates, on a long one FEXT does, and a rule of thumb learned on one does not transfer to the other. The tool says which regime you are in.
FEXT is driven by dV/dt, so halving the rise time roughly doubles it. NEXT, once saturated, does not change at all.
This catches people out because speeding up an edge is the standard fix for a timing problem. Do it on a long microstrip bus and you can double far-end crosstalk while the near-end measurement you were watching stays exactly the same. If a bus started failing after a driver strength change and NEXT looks unchanged, this is why.
The familiar guidance is to space traces three widths apart, centre to centre — two widths of gap. The tool marks that point on the curve so you can see what it buys on your stackup, which is the part the rule leaves out: coupling depends on spacing relative to the height above the plane, not relative to the trace width.
On a thin dielectric, traces can sit close together and couple very little, because each one is tightly bound to the plane beneath it. On a thick dielectric the same spacing couples strongly. That is why 3W is safe on some stackups and inadequate on others, and why the ratio the tool reports is D/H rather than S/W.
It also means the cheapest crosstalk fix is often not more spacing at all — it is moving the trace closer to its reference plane.
1/(1 + (D/H)²) form, calibrated to typical geometry. Real coupling
needs a 2D solver to extract the mutual inductance and capacitance properly, and the answers can
differ by a factor of two on unusual stackups. Use this to decide which lever to pull and roughly
how far; use SIwave or HyperLynx to sign it off.Each of these runs entirely in your browser. Nothing is uploaded, stored or sent anywhere, and none of them needs an email address.
See all 49 engineering tools →
We add tools here fairly often and write up the things worth writing up — a stackup that behaved oddly, a standard that turned out to be obsolete, a calculator that was quietly wrong. Join and you get told when something new lands.
One email a month at the very most, and usually less. No drip sequence, no sales cadence, no sharing your address with anyone. Unsubscribe whenever you like.