Loading…
Free tool · Coupling
PCB crosstalk calculator: near-end coupling and saturation length
Work out the backward crosstalk between two parallel traces, the length at which it stops growing, and how much a wider gap buys — from the pair's odd and even modes.
The short answer
Crosstalk is the signal one trace induces in another through the field they share. It is a mode problem: the pair supports an odd mode and an even mode, and the backward coefficient is (Zeven minus Zodd) over (Zeven plus Zodd). A victim at saturation sees a quarter of that coefficient. On microstrip the coupling falls roughly exponentially with gap in units of dielectric height, so a gap of three dielectric heights is usually quiet while a gap under a quarter of the height is not.
Calculator
Crosstalk and saturation length calculator
Enter the aggressor's cross-section, the edge-to-edge gap and the rise time. The tool returns the coupling coefficient, the near-end crosstalk of this run, the level it saturates at, and the length at which saturation happens. Everything comes from the pair's modal impedances, which are the same models the impedance calculator uses for a differential target.
An interface fills the rise time its edge actually has.
Line type of the aggressor. Open the button for its units, why it matters and a typical value.
Aggressor cross-section
Laminate dielectric constant (Dk). Open the button for its units, why it matters and a typical value.
Dielectric height under the aggressor. Open the button for its units, why it matters and a typical value.
Trace width. Open the button for its units, why it matters and a typical value.
Copper thickness. Open the button for its units, why it matters and a typical value.
Coupling geometry
Edge-to-edge spacing between the aggressor and the victim. Open the button for its units, why it matters and a typical value.
Parallel run length. Open the button for its units, why it matters and a typical value.
10–90% rise time of the aggressor's edge. Open the button for its units, why it matters and a typical value.ps
Frequency the impedances are evaluated at. Open the button for its units, why it matters and a typical value.GHz
Backward (near-end) crosstalk
Set by how far the odd and even modes separate, not by a lumped capacitance.
Near-end crosstalk -28.3 decibels on a 20 millimetre run.
- Near-end crosstalk
- -28.3dB
- Saturated NEXT
- -28.3dB
- Coupling k
- 15.4%
- Saturation length
- 2.5mm
- Zodd
- 48.4Ω
- Zeven
- 66.1Ω
At the saturation level.
The floor a longer run cannot go below.
(Zeven − Zodd)/(Zeven + Zodd).
tr/(2·t_d).
Past saturation
At 20.0 mm the run is past the 2.5 mm saturation length, so the crosstalk will not grow further. Only a wider gap or a different layer helps now.
Model notes
- Far-end crosstalk is not computed here. On microstrip the two modes travel at different velocities and produce a far-end term that needs the modal permittivities; on stripline it largely cancels. That extraction is the SI/PI engagement, not a closed form.
- The run is past the 2.5 mm saturation length, so the near-end crosstalk stops growing with length: shortening the parallel run no longer helps, and only opening the gap or changing the layer does.
Get an Instant QuoteEmail a layout engineer
Free, no account. The engineer reviews the inputs before a fixed quote is issued.
Want the derivation? Jump to the math behind the calculator.
The math
The formulas, with the working shown
Each equation is rendered with KaTeX, which also emits a MathML twin for screen readers and answer engines, explained in plain language, and copyable as LaTeX from the button in its corner. The variable list says what every symbol means; the model-limit note says where the formula stops being trustworthy.
Backward coupling coefficient
Coupling is how far apart the two modal impedances are, normalised by their sum.
- Z_{\text{odd}}
- odd-mode impedance of the pair
- Z_{\text{even}}
- even-mode impedance of the pair
A pair with no coupling has identical odd and even impedances and a coefficient of zero. As the gap closes, the odd mode falls and the even mode rises, and the coefficient grows.
Normalising by the sum makes it a ratio of voltages: the fraction of the aggressor's edge that appears as a backward wave at each point along the run.
Model limits
The even mode is fitted against the site's pair solve for microstrip; for stripline it is held at the uncoupled value, which overstates coupling slightly.
Near-end crosstalk at saturation
Once the run is long enough, the victim sees a quarter of the coupling coefficient.
- V_{\text{NEXT}}
- near-end crosstalk as a fraction of the aggressor swing
- k_b
- backward coupling coefficient
The quarter falls out of the coupled-line solution for a symmetric pair: backward waves from both ends add with signs that halve twice.
In decibels it is 20 log10 of that fraction, which is the number the tool reports. A coefficient of 0.15 gives about −28.5 dB.
Model limits
The quarter assumes a symmetric pair with equal terminations. An asymmetric pair, or a victim terminated differently, changes the coefficient at the ends.
Saturation length
The run stops contributing once the backward wave from the far end arrives after the edge has finished.
- t_r
- 10–90% rise time of the aggressor, ps
- t_d
- propagation delay, ps/mm
The backward wave travels at the same speed as the edge, so the round trip over a length L takes 2·t_d·L. When that exceeds the rise time, the far end of the run no longer contributes to the near end.
For a 30 ps edge on a line with 6.5 ps/mm of delay, saturation is under 2.5 mm. That is why crosstalk is a local problem: past that distance a long parallel run is no worse than a short one.
Model limits
Below saturation the crosstalk is proportional to length and inversely to rise time. That linear rise holds while the edge is much faster than the run.
How coupling falls with gap
Coupling is governed by the gap in units of dielectric height, so the height matters as much as the spacing.
- s
- edge-to-edge gap
- h
- dielectric height under the traces
The field between two traces decays with distance measured in dielectric heights, so halving the dielectric thickness doubles the effective coupling of the same gap.
That is why a thin outer dielectric is the hardest place to keep crosstalk down, and why moving a victim to a thicker inner dielectric helps even when the gap is unchanged.
Model limits
The exponential is a fit, not a derivation. It is weakest for gaps under a quarter of the dielectric height, where the tool warns and points at a pair solve.
Theory
Why coupling is a mode problem
When two traces run together, the field between them changes the impedance of both. Drive them in opposite directions and you get the odd mode, with the lower impedance; drive them together and you get the even mode, with the higher one. The difference between those two is the coupling, and every crosstalk number is built from it.
Backward crosstalk accumulates as the aggressor's edge travels: each segment injects a small wave that runs back to the victim's near end, and they add until the edge has passed. That is why the effect saturates, and why a faster edge saturates sooner in length but produces a larger coefficient.
- The backward coefficient is (Zeven − Zodd) over (Zeven + Zodd).
- At saturation a victim sees a quarter of that coefficient.
- Saturation happens at tr/(2·t_d): the backward wave has to outrun the edge.
- Coupling falls roughly exponentially with gap divided by dielectric height.
Limits
What this model does not compute
Far-end crosstalk is not modelled here. On a stripline the two modes travel at the same velocity and the far-end terms largely cancel; on a microstrip they do not, and computing the difference needs the two modal permittivities — an extraction, not a closed form. The tool says so rather than printing a number it cannot defend.
The even mode is fitted against the site's own pair solve on a 0.2 mm and a 0.3 mm microstrip, and the test suite checks the fit against that solve. For stripline the even mode is held at the uncoupled impedance, which overstates coupling slightly — the safe direction for a design check.
For a tight pair, or a channel with no margin, the real numbers come from an extraction on the actual stackup; that is the SI/PI verification service engagement.
FAQ
Common questions
It depends on the victim's noise margin, but a useful first screen is that anything worse than about −20 dB near-end deserves attention, and anything better than −30 dB is usually quiet. The number to compare against is the margin the interface's specification leaves after the channel's own loss and jitter, not a fixed rule of thumb.
Want this checked on your board?
Send the stackup and interface list. We will check the coupling against the real layout before the quote is fixed.